Skip to main content

Claude News Timeline

Stay up-to-date with the latest announcements, product updates, and news from Anthropic about Claude.

October 2025​

A statement from Dario Amodei on Anthropic's commitment to American AI leadership​

CEO Dario Amodei released a statement reaffirming Anthropic's alignment with the Trump administration on AI policy while addressing recent criticism about the company's positions. The statement highlights Anthropic's significant government partnerships including a $200 million, two-year Department of War agreement to prototype frontier AI capabilities for national security and a GSA partnership offering Claude Enterprise and Government versions at $1 pricing across federal agencies, with Claude deployed across classified networks through Palantir and Lawrence Livermore National Laboratory. Amodei emphasized Anthropic's preference for uniform federal AI standards over state-by-state regulation while defending the company's support for California's SB 53, which requires largest AI developers to disclose safety protocols while exempting companies below $500M revenue to protect startups. The company opposed a specific provision in the "One Big Beautiful Bill"—a 10-year moratorium on state AI laws—which was voted down 99-1 in the Senate. Anthropic distinguished itself as "the only frontier AI company to restrict" sales to PRC-controlled entities, forgoing revenue to prevent aiding Chinese military and intelligence operations, while addressing bias claims by citing studies from Manhattan Institute and Stanford finding Anthropic's models comparatively less politically biased than competitors.

Oct 21, 2025 | Category: Announcements

Claude Code on the web​

Anthropic introduced Claude Code on the web, a browser-based interface at claude.com/code enabling developers to delegate coding tasks directly through Anthropic-managed cloud infrastructure. The research preview allows users to launch multiple coding sessions simultaneously across different repositories from a single dashboard with real-time progress tracking, while each task operates in an isolated sandbox environment with network and filesystem restrictions. Key features include direct GitHub integration allowing developers to describe requirements while Claude handles implementation, automatic PR creation with clear change summaries, and iOS mobile support for on-the-go coding exploration. The platform uses a secure proxy service for git interactions ensuring Claude accesses only authorized repositories, with users able to define custom network policies such as permitting npm package downloads for testing and validation. Available as a research preview for Pro and Max subscribers, the announcement emphasizes this approach is "especially effective for" answering project questions, bugfixes, routine tasks, and backend changes using test-driven development, with cloud-based sessions sharing rate limits with other Claude Code usage.

Oct 20, 2025 | Category: Product

Introducing Agent Skills​

Anthropic introduced Agent Skills, a new system where Claude automatically selects and loads relevant capabilities for specific tasks from folders containing instructions, scripts, and resources. Skills are composable (can stack together), portable (same format across Claude apps, Code, and API), efficient (loads only necessary information), and powerful (includes executable code via Code Execution Tool beta). Available for Pro, Max, Team, and Enterprise users, Skills enable Claude to work with platform-specific content like Box and Notion, create professional documents including spreadsheets, presentations, and PDFs, and dramatically reduce task completion time. Early customers report significant productivity gains: Box noted "saving hours of effort," Notion experienced "less prompt wrangling on complex tasks," and Rakuten reported "what once took a day, we can now accomplish in an hour." The system includes a skill-creator skill that helps users generate custom skills, with future plans to simplify skill creation workflows and enhance enterprise-wide deployment capabilities. Anthropic cautions users to be mindful about skill sources and stick to trusted providers.

Oct 16, 2025 | Category: Product

Claude and your productivity platforms​

Anthropic announced new Microsoft 365 integration and Enterprise Search capabilities for Claude, enabling Team and Enterprise plan customers to access and analyze information across multiple productivity tools. The Microsoft 365 Connector integrates with SharePoint, OneDrive, Outlook, and Teams, allowing Claude to search and analyze documents, emails, team communications, and meeting summaries from a unified interface. Enterprise Search creates a dedicated, company-specific shared project that enables searching across multiple connected data sources simultaneously, helping retrieve comprehensive information from different organizational tools. The integration reduces time spent gathering information, facilitates faster decision-making, supports onboarding, and enables strategic analysis by providing context from various workplace tools. Example use case: "Ask Claude about your company's remote work policy and get information from HR documents, email discussions, and team guidelines in one report." Administrator enablement is required, positioning Claude as a unified, intelligent search and reasoning tool across Microsoft's productivity ecosystem.

Oct 16, 2025 | Category: Product

Introducing Claude Haiku 4.5​

Anthropic released Claude Haiku 4.5, delivering near-frontier performance with 4-5x faster speed than Sonnet 4.5 at significantly lower cost. The model achieves 90% of Sonnet 4.5's performance in agentic coding evaluations while providing similar coding performance to Claude Sonnet 4, making it ideal for real-time, low-latency tasks including chat assistants, customer service, pair programming, and computer use. Priced at $1/$5 per million input and output tokens, Haiku 4.5 is available now across Claude Code, Claude apps, the Claude API, Amazon Bedrock, and Google Cloud's Vertex AI. The model operates at AI Safety Level 2 (ASL-2) and demonstrates improved alignment over Haiku 3.5, with lower rates of misaligned behaviors compared to Sonnet 4.5 and Opus 4.1. Haiku 4.5 can be orchestrated by Sonnet 4.5 to complete complex tasks in parallel, enabling powerful multi-agent workflows for rapid prototyping, AI-assisted development, and large-scale code generation projects.

Oct 15, 2025 | Category: Product

Anthropic and Salesforce expand partnership to bring Claude to regulated industries​

Anthropic and Salesforce expanded their partnership to bring Claude to regulated industries including financial services, healthcare, cybersecurity, and life sciences through three new partnership initiatives. Claude became the first LLM provider fully integrated within Salesforce's trust boundary, serving as the preferred foundational model in Salesforce's Agentforce platform via Amazon Bedrock, enabling industry-specific AI solutions with secure data handling. The partnership includes deeper integration between Claude and Slack for conversation summarization, decision extraction, document analysis, and web searches with insights from Salesforce CRM, Tableau, and other enterprise apps, plus deployment of Claude Code across Salesforce's global engineering organization. Early customer RBC Wealth Management is using the technology to help advisors with meeting preparation, their most time-consuming task. Dario Amodei emphasized that "regulated industries need frontier AI capabilities, but they also need the appropriate safeguards," while Marc Benioff stated "we're giving every company the power to work in entirely new ways."

Oct 14, 2025 | Category: Announcements

Customize Claude Code with plugins​

Anthropic introduced the plugin system for Claude Code, enabling developers to install custom collections of slash commands, agents, MCP servers, and hooks with a single command. The system allows anyone to build and host their own plugin marketplaces using a .claude-plugin/marketplace.json file, creating a standardized extension mechanism for sharing customizations across teams and communities. Plugins can be installed via /plugin install, toggled on/off with /plugin enable/disable, and work seamlessly across both terminal and VS Code environments while reducing system prompt context when disabled. Key use cases include enforcing development standards, sharing productivity workflows, connecting internal tools, and bundling related customizations, with early community examples including Dan Ávila's marketplace featuring DevOps and documentation plugins, and Seth Hobson's GitHub repo containing over 80 specialized sub-agents.

Oct 9, 2025 | Category: Product

Deloitte will make Claude available to 470,000 people across its global network​

Anthropic and Deloitte announced an expanded alliance making Claude available to more than 470,000 Deloitte people across their global network, representing Anthropic's largest enterprise AI deployment to date. The collaboration includes establishing a Claude Center of Excellence with trained specialists to develop implementation frameworks and provide technical support for scaling AI pilots to production. Deloitte and Anthropic are co-creating a formal certification program to train and certify 15,000 professionals on Claude to support implementations across Deloitte's network and internal AI transformation efforts. The partnership focuses on building industry-specific AI solutions for regulated industries like financial services, healthcare and life sciences, and public services, combining Claude's safety-first design with Deloitte's Trustworthy AI™ framework to drive transparency in model decision-making.

Oct 6, 2025 | Category: Announcements

Building AI for cyber defenders​

Anthropic focused on improving Claude's cybersecurity capabilities for defensive purposes with Claude Sonnet 4.5, deliberately avoiding offensive capability development. The model succeeded on 76.5% of Cybench challenges with 10 attempts and discovered new vulnerabilities in 5% of targets (up to 33% with 30 trials), while demonstrating emerging capability in patch generation. Collaboration with HackerOne and CrowdStrike showed organizations reduced vulnerability intake time by 44% and improved accuracy by 25% when applying the model to vulnerability remediation and threat analysis. Key use cases include Security Operations Center (SOC) automation, SIEM analysis, secure network engineering, and active defense strategies, with Anthropic emphasizing that "we should not cede the cyber advantage derived from AI to attackers and criminals."

Oct 3, 2025 | Category: Policy

Claude and Slack​

Anthropic announced an enhanced integration between Claude and Slack, enabling teams to access AI capabilities directly within their communication workspace through two methods: adding Claude directly to Slack or connecting Slack to Claude apps for message search and reference. The integration allows Claude to pull context from docs, emails, connected tools, and Slack channels to provide intelligent assistance for tasks like meeting preparation, project coordination, onboarding support, and documentation creation. Users can interact with Claude through direct messages, the AI assistant panel, or by mentioning @Claude in threads, with all interactions respecting existing Slack permissions and Claude security settings. The feature is available through the Slack Marketplace for customers with paid Slack plans and requires Claude Team or Enterprise subscriptions, bridging AI capabilities directly into work communication platforms for seamless, intelligent workflows.

Oct 1, 2025 | Category: Product

How enterprises are driving AI transformation with Claude​

Anthropic showcased enterprise AI transformation success stories demonstrating significant business impact across multiple industries. Novo Nordisk developed the NovoScribe AI platform reducing clinical documentation time from 10+ weeks to 10 minutes (90% reduction) and review cycles by 50% with an 11-person development team. Cox Automotive integrated Claude across CRM and dealer platforms, more than doubling consumer lead responses and test drive appointments with 80% positive feedback on AI-generated vehicle listings and over 9,000 client deliverables generated. Palo Alto Networks accelerated software development with 20-30% increased feature velocity, reduced developer onboarding from months to weeks while onboarding 2,500 developers (planning 3,500), and enabled junior developers to complete integration tasks 70% faster. Salesforce created autonomous AI agents through Einstein 1 Studio for end-to-end workflow orchestration, while IG Group automated analytics workflows saving 70 hours weekly, doubling productivity in certain use cases, and achieving full ROI within three months. Common success patterns include starting with concrete business problems, investing in people and training, measuring tangible metrics, and building for scale from day one.

Oct 1, 2025 | Category: Announcements

September 2025​

Introducing Claude Sonnet 4.5​

Anthropic released Claude Sonnet 4.5, described as "the Anthropic's best coding model" with significant advances in computer use capabilities, reasoning, and mathematical performance. The model leads SWE-bench Verified evaluation for software coding, can maintain focus for over 30 hours on complex tasks, and shows enhanced performance across domains like finance, law, medicine, and STEM. Key features include improved multi-step task handling, better code generation, and leading the OSWorld benchmark at 61.4%. The model comes with upgraded Claude Code features including checkpoints for saving progress, a refreshed terminal interface, and a native VS Code extension. Additionally, context editing and memory tools are now available, along with code execution capabilities in Claude apps and a new Claude for Chrome extension. Despite being the "most aligned frontier model" with reduced sycophancy and deception, pricing remains the same as Claude Sonnet 4 at $3/$15 per million tokens.

Sep 29, 2025 | Category: Product

Managing context on the Claude Developer Platform​

Anthropic introduced powerful context management capabilities to help developers build AI agents capable of handling long-running tasks at higher performance without hitting context limits. The new features include Context Editing that automatically removes stale tool calls and results from the context window, increasing model performance by focusing on relevant context, and a Memory Tool that enables storing and retrieving information outside the context window using a file-based system that persists across conversations. Performance improvements are significant: combining memory tool and context editing improved performance by 39% in agentic search evaluations, while context editing alone delivered 29% improvement and reduced token consumption by 84% in 100-turn web searches. The features are available in public beta on the Claude Developer Platform and natively available in Amazon Bedrock and Google Cloud's Vertex AI, introduced with the Claude Sonnet 4.5 model.

Sep 29, 2025 | Category: Product

Enabling Claude Code to work more autonomously​

Anthropic announced major upgrades to Claude Code enabling more autonomous development workflows, powered by Claude Sonnet 4.5. Key features include a native VS Code Extension (Beta) with real-time change tracking and inline code difference visualization, an enhanced terminal experience with improved status visibility and searchable prompt history (Ctrl+r), and advanced autonomous capabilities through a checkpointing system that allows saving code state and instant reversion to previous versions. The upgrade introduces sophisticated agent features including subagents for parallel tasks, hooks for automatic actions, and background task support. New SDK capabilities support custom agent creation with specialized workflow agents and provide context management and permissions frameworks. These upgrades allow developers to "confidently delegate broad tasks like extensive refactors or feature exploration to Claude Code," making it more autonomous, flexible, and capable of handling increasingly complex development tasks.

Sep 29, 2025 | Category: Product

Anthropic expands global leadership in enterprise AI, naming Chris Ciauri as Managing Director of International​

Anthropic appointed Chris Ciauri as Managing Director of International, bringing over 25 years of enterprise technology experience from leadership roles at Google Cloud, Salesforce, and Unily. Ciauri previously served as president of EMEA at Google Cloud and scaled Salesforce's EMEA business from $200 million to over $3 billion. The appointment supports Anthropic's rapid international expansion, with enterprise customers growing from 1,000 to 300,000 in two years, and plans for over 100 new roles across Dublin, London, and Zurich offices, plus their first Asia office in Tokyo.

Sep 26, 2025 | Category: Announcements

Claude is now available in Microsoft 365 Copilot​

Anthropic announced that Claude models (Sonnet 4 and Opus 4.1) are now available in Microsoft 365 Copilot, expanding AI model choices for enterprise customers through two key integrations: the Researcher Agent powered by Claude Opus 4.1 for complex multi-step research across work content like emails, chats, and files, and Copilot Studio for creating and customizing enterprise-grade AI agents using Anthropic models. The rollout is happening through Microsoft's Frontier Program, with administrators able to enable access in the Microsoft 365 admin center.

Sep 24, 2025 | Category: Announcements

Anthropic Economic Index report: Uneven geographic and enterprise AI adoption​

Anthropic released its third Economic Index report analyzing global AI adoption patterns, revealing significant geographic concentration and enterprise automation trends. The study shows AI usage is strongly correlated with income per capita, with technologically advanced countries like Israel and Singapore leading adoption rates while emerging economies show minimal usage. The report introduces the Anthropic AI Usage Index (AUI) to measure population-adjusted usage and finds that higher-adoption countries tend toward collaborative augmentation while lower-adoption regions prefer task automation. Enterprise API customers show 77% automation patterns compared to 50% automation on Claude.ai, suggesting significant labor market implications as businesses increasingly delegate complete tasks to AI systems.

Sep 15, 2025 | Category: Societal Impacts

Anthropic Economic Index: Tracking AI's role in the US and global economy​

Anthropic expanded its Economic Index to include detailed geographic analysis showing how local economies shape AI usage patterns across countries and US states. Within the US, Washington DC leads per-capita usage with specialized focus on document editing and job applications, while states like California show disproportionate IT-related requests reflecting their tech sector concentration. The analysis reveals that directive automation has increased sharply from 27% to 39% over nine months, suggesting growing user confidence in AI capabilities. Countries with higher per-capita usage demonstrate more diverse task portfolios beyond coding, while emerging markets concentrate primarily on software development tasks.

Sep 15, 2025 | Category: Societal Impacts

Strengthening our safeguards through collaboration with US CAISI and UK AISI​

Anthropic announced ongoing collaboration with the US Center for AI Standards and Innovation (CAISI) and UK AI Security Institute (AISI) to strengthen AI safety measures through government red-teaming. The partnership involves providing government teams access to Constitutional Classifiers and models like Claude Opus 4 during development phases to identify vulnerabilities before deployment. Key discoveries include prompt injection vulnerabilities, cipher-based attacks, and universal jailbreaks that enabled targeted improvements to safeguard architectures. The collaboration demonstrates effective public-private partnership models where government expertise in cybersecurity and threat modeling complements industry AI development.

Sep 12, 2025 | Category: Announcements

Bringing memory to teams at work​

Claude introduces memory capabilities that remember you and your team's projects, preferences, and work patterns, eliminating the need to re-explain context in ongoing conversations. Memory creates separate memories for each project, ensuring confidential discussions remain isolated while maintaining continuity across complex initiatives. The feature includes Incognito chat for sensitive conversations that don't save to memory or history, and granular user controls allowing you to view and edit exactly what Claude remembers.

Sep 11, 2025 | Category: Product

Claude can now create and edit files​

Claude can now create and edit Excel spreadsheets, documents, PowerPoint slides, and PDFs directly in Claude.ai and the desktop app, transforming how users work by producing ready-to-use files from instructions and data. This feature is available as a preview for Max, Team, and Enterprise users, with Pro users getting access in coming weeks.

Sep 9, 2025 | Category: Product

Anthropic is endorsing SB 53​

Anthropic is endorsing California's SB 53, a bill governing powerful AI systems that requires companies to develop safety frameworks, publish transparency reports, and report critical incidents to the state. The company supports this trust but verify approach as it formalizes practices already followed by frontier AI companies while creating a level playing field for safety disclosure.

Sep 8, 2025 | Category: Announcements

Updating restrictions of sales to unsupported regions​

Anthropic is strengthening regional restrictions to prevent companies controlled by adversarial nations like China from accessing their services, even through subsidiaries in other countries. The updated policy prohibits entities more than 50% owned by companies from unsupported regions, addressing national security risks where authoritarian governments can compel data sharing and intelligence cooperation.

Sep 4, 2025 | Category: Announcements

Anthropic raises $13B Series F at $183B post-money valuation​

Anthropic completed a $13 billion Series F funding round led by ICONIQ, valuing the company at $183 billion post-money with participation from major investors including Fidelity, Lightspeed, and BlackRock. The funding reflects Anthropic's rapid growth from $1 billion run-rate revenue in early 2025 to over $5 billion by August, making it one of the fastest-growing technology companies in history.

Sep 2, 2025 | Category: Announcements

August 2025​

Updates to Consumer Terms and Privacy Policy​

Anthropic updated its Consumer Terms and Privacy Policy to allow training new models using data from Free, Pro, and Max accounts when users opt-in, with an extended 5-year data retention period for those who consent. Current users have until September 28, 2025 to make their selection, while new users choose during signup.

Aug 28, 2025 | Category: Product

Introducing the Anthropic National Security and Public Sector Advisory Council​

Anthropic formed the National Security and Public Sector Advisory Council, featuring bipartisan former government leaders including Senators Roy Blunt and Jon Tester, former Defense officials, and intelligence community veterans. The council will help identify high-impact AI applications for national security, develop industry standards, and strengthen public-private partnerships.

Aug 27, 2025 | Category: Announcements

Detecting and countering misuse of AI: August 2025​

Anthropic's threat intelligence report reveals sophisticated AI misuse including a large-scale extortion operation using Claude Code, North Korean employment fraud schemes, and AI-generated ransomware sales. The report shows cybercriminals are weaponizing agentic AI to perform attacks rather than just advise, lowering barriers to sophisticated cybercrime for actors with minimal technical skills.

Aug 27, 2025 | Category: Announcements

Claude Code and new admin controls for business plans​

Claude Code is now available for Team and Enterprise customers through premium seats that include enhanced usage and coding capabilities, allowing seamless transitions between ideation and implementation. New admin controls include self-serve seat management, granular spend controls, usage analytics, and a Compliance API for real-time programmatic access to usage data.

Aug 20, 2025 | Category: Product

Anthropic appoints Hidetoshi Tojo as Head of Japan and announces hiring plans​

Anthropic appointed Hidetoshi Tojo as Head of Japan, bringing extensive experience from scaling technology companies including Snowflake, Google Cloud, and Microsoft across Japan operations. The appointment supports Anthropic's expansion plans including opening their first Asia office in Tokyo and hiring local talent to serve Japanese customers.

Aug 6, 2025 | Category: Announcements

Automate security reviews with Claude Code​

Claude Code introduced automated security reviews through a new /security-review command and GitHub Actions integration that identifies vulnerabilities like SQL injection, XSS, and authentication flaws before code reaches production. The features allow ad-hoc security analysis from terminals and automatic pull request reviews with inline comments and fix recommendations.

Aug 6, 2025 | Category: Announcements

July 2025​

Anthropic to sign the EU Code of Practice​

Anthropic announced its intention to sign the EU's General-Purpose AI Code of Practice, viewing it as advancing transparency, safety, and accountability principles for frontier AI development. The Code establishes mandatory Safety and Security Frameworks building on Anthropic's Responsible Scaling Policy while maintaining flexibility to adapt with evolving technology.

Jul 21, 2025 | Category: Policy

Paul Smith to join Anthropic as Chief Commercial Officer​

Paul Smith will join Anthropic as its first Chief Commercial Officer later in 2025, bringing over 30 years of experience building global go-to-market organizations at Microsoft, Salesforce, and ServiceNow. The appointment comes during exceptional growth with hundreds of thousands of active API customers and Claude Code revenue growing over 5x in two months.

Jul 15, 2025 | Category: Announcements

Anthropic and the Department of Defense to advance responsible AI in defense operations​

Anthropic received a $200 million, two-year agreement from the U.S. Department of Defense to prototype frontier AI capabilities for national security applications. The partnership will focus on developing working prototypes fine-tuned on DOD data, collaborating on adversarial AI mitigation, and accelerating responsible AI adoption across defense operations.

Jul 14, 2025 | Category: Announcements

June 2025​

National security expert Richard Fontaine appointed to Anthropic’s long-term benefit trust​

Richard Fontaine, CEO of the Center for a New American Security, was appointed to Anthropic's Long-Term Benefit Trust, bringing extensive national security experience including roles at the NSC, State Department, and Defense Policy Board. His appointment addresses the growing intersection between AI capabilities and geopolitical risks, strengthening the Trust's ability to guide Anthropic through complex decisions.

Jun 7, 2025 | Category: Announcements

May 2025​

Reed Hastings appointed to Anthropic’s board of directors​

Reed Hastings, Netflix co-founder and former CEO, was appointed to Anthropic's board of directors by the Long-Term Benefit Trust. Hastings brings 25+ years of scaling experience and recently donated $50 million to establish an AI and Humanity research initiative at Bowdoin College.

May 28, 2025 | Category: Announcements

New capabilities for building agents on the Anthropic API​

Anthropic announced four new API capabilities for building AI agents: code execution tool for Python analysis, MCP connector for external system integration, Files API for document management, and extended prompt caching up to one hour. These features enable developers to build more powerful agents that can execute code, connect to third-party services, manage files across sessions, and maintain context cost-effectively.

May 22, 2025 | Category: Product

April 2025​

Claude takes research to new places​

Anthropic introduced Research capabilities that enable Claude to conduct multi-step web searches and a Google Workspace integration connecting Gmail, Calendar, and Google Docs. Research allows Claude to operate agentically, exploring different angles and building comprehensive answers with citations, while the Google Workspace integration provides deeper work context without manual file uploads.

Apr 15, 2025 | Category: Product

Anthropic appoints Guillaume Princen as Head of EMEA and announces 100+ new roles across the region​

Guillaume Princen was appointed as Anthropic's Head of EMEA, bringing experience from scaling Stripe's European operations across 12 offices and serving as CEO of Mooncard. Anthropic plans to create over 100 new roles across sales, engineering, research, and operations in Dublin and London offices this year.

Apr 8, 2025 | Category: Announcements

Introducing Anthropic's first developer conference: Code with Claude​

Anthropic announced Code with Claude, its first developer conference taking place May 22, 2025 in San Francisco at The Midway. The hands-on event will focus on real-world implementations using the Anthropic API, CLI tools, and Model Context Protocol (MCP), featuring interactive workshops, product roadmap discussions, and networking opportunities.

Apr 3, 2025 | Category: Event

March 2025​

Progress from our Frontier Red Team​

Anthropic's Frontier Red Team reported significant progress in AI capabilities across cybersecurity and biology, with Claude advancing from high school to undergraduate level in cybersecurity CTF challenges within one year. While models show early warning signs in dual-use capabilities and exceed expert baselines in some biology tasks, they still fall short of thresholds for substantially elevated national security risks.

Mar 19, 2025 | Category: Policy

February 2025​

Anthropic partners with U.S. National Labs for first 1,000 Scientist AI Jam​

Anthropic partnered with the U.S. Department of Energy for the first 1,000 Scientist AI Jam, bringing together scientists from multiple National Laboratories to evaluate Claude 3.7 Sonnet on scientific research and national security applications. The event will test Claude's capabilities across real-world research problems including hypothesis generation, experiment planning, and result analysis.

Feb 28, 2025 | Category: Announcements

Introducing Anthropic's Transparency Hub​

Anthropic launched its Transparency Hub, providing detailed information on safety protocols, risk mitigation strategies, platform abuse detection, and governance policies. The hub includes the first periodic report on key metrics like banned accounts, appeals, and government requests, designed as a unified framework to address diverse transparency requirements.

Feb 27, 2025 | Category: Societal Impacts

Claude and Alexa+​

Claude models now power Amazon's Alexa+ through a collaboration between Anthropic and Amazon teams, with Anthropic's Chief Product Officer Mike Krieger leading the integration effort. Alexa+ benefits from Claude's advanced capabilities and safety features, including jailbreaking resistance, and accesses Claude through Amazon Bedrock.

Feb 26, 2025 | Category: Announcements

Statement from Dario Amodei on the Paris AI Action Summit​

Dario Amodei issued a statement following the Paris AI Action Summit emphasizing three critical areas needing greater focus: ensuring democratic societies lead in AI development, addressing growing security risks including CBRN threats and autonomous AI dangers, and preparing for massive economic disruption from AI advancement. He warned that by 2026-2027, AI capabilities could resemble a country of geniuses in a datacenter and called for accelerated action to match the pace of AI progress.

Feb 11, 2025 | Category: Announcements

Lyft to bring Claude to more than 40 million riders and over 1 million drivers​

Lyft is partnering with Anthropic to integrate Claude across its platform serving over 40 million riders and 1 million drivers, focusing on AI-powered solutions, early testing of new models, and engineering advancement. The collaboration has already shown significant impact, with Lyft's Claude-powered customer care assistant reducing resolution time by 87% while handling thousands of daily inquiries.

Feb 6, 2025 | Category: Announcements

January 2025​

Introducing Citations on the Anthropic API​

Anthropic launched Citations, a new API feature that allows Claude to ground its responses in source documents by providing detailed references to exact sentences and passages used to generate answers. The feature is generally available on the Anthropic API and Google Cloud's Vertex AI, with internal evaluations showing up to 15% improvement in recall accuracy compared to custom implementations.

Jan 23, 2025 | Category: Product

Anthropic achieves ISO 42001 certification for responsible AI​

Anthropic achieved ISO 42001 certification, the first international standard for AI governance and management systems, making it one of the first frontier AI labs to receive this accreditation. The certification validates Anthropic's comprehensive framework for identifying, assessing, and mitigating AI risks through policies, testing, transparency measures, and established oversight responsibilities.

Jan 13, 2025 | Category: Announcements

December 2024​

Elections and AI in 2024: observations and learnings​

Anthropic published observations from the 2024 election cycle, noting that election-related activity constituted less than 0.5% of overall Claude usage, rising to just over 1% during peak election weeks. The company implemented comprehensive safety measures including policy enforcement, rigorous testing, and directing users to authoritative sources, while using their new Clio tool to analyze usage patterns and ensure responsible AI deployment during elections.

Dec 12, 2024 | Category: Societal Impacts

Claude 3.5 Haiku on AWS Trainium2 and model distillation in Amazon Bedrock​

Anthropic announced Claude 3.5 Haiku optimization for AWS Trainium2 chips, delivering up to 60% faster inference speeds, and introduced model distillation in Amazon Bedrock that enables Claude 3 Haiku to achieve Claude 3.5 Sonnet-like accuracy for specific tasks. The company also reduced Claude 3.5 Haiku pricing to $0.80 per million input tokens and $4 per million output tokens across all platforms.

Dec 3, 2024 | Category: Product

November 2024​

Tailor Claude’s responses to your personal style​

Anthropic introduced custom styles for Claude.ai, allowing users to tailor Claude's responses to match their communication preferences, tone, and structure. Users can choose from preset options (Formal, Concise, Explanatory) or generate custom styles by uploading sample content, enabling Claude to adapt to individual workflows and writing preferences.

Nov 26, 2024 | Category: Product

Introducing the Model Context Protocol​

Anthropic open-sourced the Model Context Protocol (MCP), a new standard for connecting AI assistants to data sources including content repositories, business tools, and development environments. MCP provides a universal protocol to replace fragmented integrations, with local server support in Claude Desktop apps and pre-built servers for popular enterprise systems like Google Drive, Slack, GitHub, and Postgres.

Nov 25, 2024 | Category: Announcements

Powering the next generation of AI development with AWS​

Anthropic announced an expanded $4 billion partnership with AWS, bringing Amazon's total investment to $8 billion while establishing AWS as their primary cloud and training partner. The collaboration includes deep technical work on AWS Trainium accelerators, with Anthropic contributing to hardware optimization and the AWS Neuron software stack to train their most advanced foundation models.

Nov 22, 2024 | Category: Announcements

Improve your prompts in the developer console​

Anthropic introduced a prompt improver feature in the developer console that automatically refines existing prompts using advanced engineering techniques like chain-of-thought reasoning and example standardization. The tool also includes structured example management and evaluation capabilities, with testing showing 30% accuracy improvements for classification tasks and 100% word count adherence for summarization.

Nov 14, 2024 | Category: Product

October 2024​

The case for targeted regulation​

Anthropic published their position on AI regulation, advocating for targeted, RSP-based (Responsible Scaling Policy) regulation that focuses on transparency, incentivizing better safety practices, and maintaining simplicity. The company argues that governments should urgently act within 18 months to implement proportionate regulation that addresses catastrophic risks while supporting innovation, emphasizing the need for surgical rather than broad regulatory approaches.

Oct 31, 2024 | Category: Policy

Claude 3.5 Sonnet on GitHub Copilot​

Claude 3.5 Sonnet became available on GitHub Copilot, bringing Claude's coding capabilities to over 100 million developers through Visual Studio Code and GitHub.com. The upgraded model outperforms all publicly available models on SWE-bench Verified and achieves 93.7% on HumanEval, enabling developers to write production-ready code, debug with inline chat, and create comprehensive test suites.

Oct 29, 2024 | Category: Announcements

Introducing the analysis tool in Claude.ai​

Anthropic launched the analysis tool in Claude.ai, a built-in JavaScript code sandbox that enables Claude to write and run code for data processing, analysis, and real-time insights. The tool allows Claude to work like a data analyst by systematically processing data from CSV files, providing mathematically precise and reproducible answers for various business teams including marketing, sales, product management, and finance.

Oct 24, 2024 | Category: Product

Developing a computer use model​

Anthropic announced Claude's new computer use capability, allowing the AI to control computers by taking screenshots, moving cursors, clicking, and typing like humans do. This represents a significant breakthrough in AI progress, enabling Claude to use any software directly rather than requiring custom tools, though the feature remains slow and error-prone in its current beta state.

Oct 22, 2024 | Category: Announcements

Announcing our updated Responsible Scaling Policy​

Anthropic released a major update to its Responsible Scaling Policy (RSP), introducing more flexible risk assessment methods and refined capability thresholds for upgrading safety measures. The updated framework includes new AI Safety Level Standards and specific thresholds for autonomous AI research capabilities and CBRN weapons assistance, while maintaining the core commitment not to deploy models without adequate safeguards.

Oct 15, 2024 | Category: Announcements

U.S. Elections Readiness​

Anthropic outlined its comprehensive election readiness measures for the 2024 U.S. elections, including updated usage policies prohibiting campaigning and misinformation, automated enforcement systems, and redirecting users to authoritative voting information. The company implemented strict controls against election-related misuse while conducting ongoing vulnerability testing and policy refinements to ensure responsible AI use during the election period.

Oct 8, 2024 | Category: Societal Impacts

Introducing the Message Batches API​

Anthropic launched the Message Batches API, enabling developers to process up to 10,000 queries asynchronously at 50% lower cost than standard API calls. The service processes batches within 24 hours and offers enhanced throughput without impacting standard rate limits, making large-scale data processing more economically viable for tasks like dataset analysis and document processing.

Oct 8, 2024 | Category: Product

September 2024​

Fine-tuning for Claude 3 Haiku in Amazon Bedrock is now generally available​

Fine-tuning for Claude 3 Haiku became generally available in Amazon Bedrock, allowing developers to customize the fastest and most cost-effective Claude model with their own data. The feature enables enhanced domain expertise, cost savings, and faster response times, with demonstrated improvements showing fine-tuned Haiku outperforming even Claude 3.5 Sonnet base model in specialized financial analysis tasks.

Sep 23, 2024 | Category: Product

Introducing Contextual Retrieval​

Anthropic introduced Contextual Retrieval, a new technique that dramatically improves RAG (Retrieval-Augmented Generation) systems by adding chunk-specific context before embedding. The method reduces retrieval failures by 49% when combined with BM25, and up to 67% when paired with reranking, significantly enhancing the accuracy of knowledge base searches for AI applications.

Sep 19, 2024 | Category: Product

Workspaces in the Anthropic API Console​

Anthropic launched Workspaces in the API Console to help developers manage multiple Claude deployments across different environments and use cases. The feature provides granular spend limits, resource organization, independent rate limit management, streamlined access controls, and workspace-level usage monitoring for better project management and security.

Sep 10, 2024 | Category: Product

Claude for Enterprise​

Anthropic announced Claude for Enterprise, offering a 500K context window, enhanced security features like SSO and role-based permissions, and native GitHub integration for codebase collaboration. The enterprise plan enables organizations to securely work with internal knowledge while maintaining data protection, with early customers like GitLab and Midjourney using it across various business functions.

Sep 4, 2024 | Category: Product

Salesforce teams up with Anthropic to enhance Einstein capabilities with Claude​

Salesforce partnered with Anthropic to integrate Claude models (3.5 Sonnet, Opus, and Haiku) into Einstein capabilities through Amazon Bedrock. The integration allows Salesforce customers to use Claude for sales, marketing, customer service, and other enterprise functions while maintaining security through Salesforce's Einstein Trust Layer and existing compliance standards.

Sep 3, 2024 | Category: Announcements

August 2024​

Artifacts are now generally available​

Artifacts became generally available to all Claude.ai users across Free, Pro, and Team plans, including mobile iOS and Android apps. The feature creates a dedicated workspace for users to instantly see, iterate, and build on creative work with Claude, from code snippets and visualizations to interactive prototypes, with tens of millions of Artifacts created since the preview launch.

Aug 27, 2024 | Category: Announcements

Prompt caching with Claude​

Anthropic introduced prompt caching for Claude, enabling developers to cache frequently used context between API calls with up to 90% cost reduction and 85% latency improvement. The feature is particularly effective for conversational agents, coding assistants, and large document processing, with cached content costing only 10% of base input token prices while cache writes cost 25% more.

Aug 14, 2024 | Category: Product

Expanding our model safety bug bounty program​

Anthropic expanded its model safety bug bounty program to focus on identifying universal jailbreak attacks that could bypass AI safety guardrails across high-risk domains like CBRN and cybersecurity. The invite-only program offers rewards up to $15,000 for novel vulnerabilities and provides early access to test next-generation safety mitigation systems before public deployment.

Aug 8, 2024 | Category: Announcements

Claude is now available in Brazil​

Anthropic announced Claude's availability in Brazil on August 1, 2024, making the AI assistant accessible through Claude.ai, mobile apps, and the API. Brazilian users can access Claude for free or subscribe to Pro (R$ 110/month) and Team plans (R$ 165/month) with enhanced usage limits and features.

Aug 1, 2024 | Category: Announcements

July 2024​

Anthropic partners with Menlo Ventures to launch Anthology Fund​

Anthropic partnered with Menlo Ventures to launch the $100 million Anthology Fund on July 17, 2024, aimed at supporting startups building with Anthropic technology. The fund focuses on five key areas including AI infrastructure, healthcare applications, consumer AI solutions, and trust and safety tooling.

Jul 17, 2024 | Category: Announcements

Claude Android app​

Anthropic launched the Claude Android app on July 16, 2024, bringing Claude 3.5 Sonnet to Android users for free. The app offers cross-platform conversation continuity with web and iOS versions, vision capabilities for image analysis, multilingual processing, and advanced reasoning for complex tasks.

Jul 16, 2024 | Category: Product

Fine-tune Claude 3 Haiku in Amazon Bedrock​

Anthropic announced fine-tuning capabilities for Claude 3 Haiku in Amazon Bedrock on July 11, 2024, allowing customers to customize the model for specialized business tasks. Fine-tuning enables better performance on domain-specific tasks, faster speeds at lower costs, and consistent brand-aligned formatting.

Jul 11, 2024 | Category: Product

Evaluate prompts in the developer console​

Anthropic introduced prompt evaluation features in the developer console on July 9, 2024, streamlining AI application development. The new tools include automatic prompt generation powered by Claude 3.5 Sonnet, test case generation, and side-by-side output comparison capabilities.

Jul 9, 2024 | Category: Product

A new initiative for developing third-party model evaluations​

Anthropic launched a new initiative on July 1, 2024, to fund third-party organizations developing AI model evaluations, addressing the limited evaluation ecosystem. The initiative prioritizes AI Safety Level assessments, advanced capability metrics, and evaluation infrastructure development.

Jul 1, 2024 | Category: Announcements

June 2024​

Expanding access to Claude for government​

Anthropic expanded Claude access to government users on June 26, 2024, making Claude 3 Haiku and Sonnet available through AWS Marketplace for the US Intelligence Community and AWS GovCloud. The expansion includes carefully crafted contractual exceptions for legally authorized foreign intelligence analysis while maintaining restrictions on disinformation, weapons design, and malicious cyber operations.

Jun 26, 2024 | Category: Announcements

Collaborate with Claude on Projects​

Anthropic launched Projects on Claude.ai for Pro and Team users on June 25, 2024, enabling organized collaboration with curated knowledge sets and shared conversations. Projects include a 200K context window for documents and code, custom instructions for tailored responses, and team sharing capabilities.

Jun 25, 2024 | Category: Product

Claude 3.5 Sonnet​

Anthropic released Claude 3.5 Sonnet on June 21, 2024, setting new industry benchmarks while operating at twice the speed of Claude 3 Opus. The model excels in graduate-level reasoning, coding proficiency (solving 64% of problems in internal evaluations), and vision capabilities for chart interpretation and text transcription.

Jun 21, 2024 | Category: Announcements

Challenges in red teaming AI systems​

Anthropic published insights on red teaming AI systems on June 12, 2024, detailing various approaches including expert domain testing, automated red teaming, and multimodal evaluations. The post outlines challenges in standardizing red teaming practices and proposes policy recommendations including funding NIST for technical standards and supporting independent testing organizations.

Jun 12, 2024 | Category: Policy

Anthropic detailed its approach to testing and mitigating elections-related risks on June 6, 2024, combining Policy Vulnerability Testing with external experts and automated evaluations. The company implemented multiple mitigations including system prompt updates, model fine-tuning, and policy refinements to improve accuracy and appropriate referrals to authoritative sources.

Jun 6, 2024 | Category: Policy

Introducing Claude to Canada​

Anthropic introduced Claude to Canada on June 5, 2024, making the AI assistant available through Claude.ai, iOS app, API, and Team plan. Canadian users can access Claude for free or subscribe to Pro for CA$28/month and Team for CA$42/month per user, with access to all Claude 3 models and enhanced usage limits.

Jun 5, 2024 | Category: Announcements

May 2024​

Claude can now use tools​

Claude's tool use capabilities became generally available across the entire Claude 3 model family on May 30, 2024, enabling Claude to interact with external tools and APIs for tasks like data extraction, API calls, and database searches. This feature includes streaming support, forced tool selection, and image compatibility, significantly expanding Claude's practical applications for developers and businesses.

May 30, 2024 | Category: Product

Jay Kreps appointed to Anthropic's Board of Directors​

Jay Kreps, co-founder and CEO of Confluent, joined Anthropic's Board of Directors on May 29, 2024, bringing extensive experience in building and scaling tech companies and expertise in data infrastructure. His appointment by the Long-Term Benefit Trust comes as Anthropic prepares for its next phase of growth, while Luke Muehlhauser stepped down from the board to focus on his work at Open Philanthropy.

May 29, 2024 | Category: Announcements

Golden Gate Claude​

Golden Gate Claude was a 24-hour research demonstration released on May 23, 2024, showcasing Anthropic's breakthrough in AI interpretability by artificially amplifying Claude's Golden Gate Bridge feature to make it obsessively reference the bridge in all responses. This demonstrated the ability to surgically modify specific neural pathways in AI models, representing a significant advance in understanding and potentially controlling AI behavior for safety purposes.

May 23, 2024 | Category: Product

Krishna Rao joins Anthropic as Chief Financial Officer​

Krishna Rao joined Anthropic as Chief Financial Officer on May 21, 2024, bringing nearly 20 years of strategic finance experience from companies like Fanatics Commerce, Cedar, and Airbnb where he helped navigate the pandemic and IPO. His expertise in financial strategy, capital allocation, and scaling high-growth organizations will support Anthropic's enterprise momentum and international expansion plans.

May 21, 2024 | Category: Announcements

Generate better prompts in the developer console​

Anthropic launched a prompt generator feature in the developer console on May 20, 2024, that automatically creates production-ready prompt templates using best practices like chain-of-thought reasoning and role setting. The tool helps both new and experienced prompt engineers by generating effective, precise prompts based on task descriptions, significantly reducing development time and improving output quality.

May 20, 2024 | Category: Product

Reflections on our Responsible Scaling Policy​

Anthropic published reflections on implementing their Responsible Scaling Policy on May 20, 2024, outlining five key commitments including establishing Red Line Capabilities, conducting frontier risk evaluations, and developing ASL-3 safety standards. The post detailed lessons learned from operationalizing the policy, including challenges in threat modeling, evaluation methodologies, and the need for significant organizational investment in safety measures.

May 20, 2024 | Category: Policy

Mike Krieger joins Anthropic as Chief Product Officer​

Mike Krieger, co-founder and former CTO of Instagram, joined Anthropic as Chief Product Officer on May 15, 2024, bringing experience scaling platforms to over a billion users and building Artifact, a personalized news app. He will oversee product engineering, management, and design efforts as Anthropic expands its enterprise applications and brings Claude to a wider audience.

May 15, 2024 | Category: Announcements

Claude is now available in Europe​

Claude became available across Europe on May 14, 2024, including Claude.ai, the iOS app, and the Team plan, following the earlier European launch of the Claude API. The expansion provides Europeans access to Claude's strong multilingual capabilities in French, German, Spanish, and Italian, with Claude Pro available for €18 + VAT per month and Team plan for €28 + VAT per user per month.

May 14, 2024 | Category: Announcements

Updating our Usage Policy​

Anthropic updated its Usage Policy (formerly Acceptable Use Policy) on May 10, 2024, with changes taking effect June 6, 2024, streamlining guidelines into Universal Usage Standards and clarifying restrictions on election interference and misinformation. The update also added requirements for high-risk use cases like healthcare decisions, expanded access for minors through organizations with safety measures, and strengthened privacy protections against biometric analysis and government censorship.

May 10, 2024 | Category: Announcements

Introducing the Claude Team plan and iOS app​

Anthropic launched the Claude Team plan ($30 per user per month, minimum 5 seats) and iOS app on May 1, 2024, providing teams with increased usage, access to the full Claude 3 model family, and administrative tools. The free iOS app offers seamless syncing with web chats, vision capabilities for photo analysis, and mobile access to Claude's frontier intelligence capabilities.

May 1, 2024 | Category: Product

April 2024​

Aligning on child safety principles​

Anthropic committed to child safety principles on April 23, 2024, as part of a Safety by Design initiative led by Thorn and All Tech Is Human to prevent AI-generated child sexual abuse material (AIG-CSAM). The commitment includes specific measures across development, deployment, and maintenance phases, such as responsibly sourcing training data, detecting abusive content, and reporting violations to NCMEC.

Apr 23, 2024 | Category: Announcements

March 2024​

Third-party testing as a key ingredient of AI policy​

Anthropic published a policy position on March 25, 2024, advocating for third-party testing as essential for AI policy, arguing that frontier AI systems need independent oversight to validate safety claims and prevent misuse. The company outlined how such testing regimes should work, including automated initial screening followed by expert human evaluation, and emphasized the need for government funding of testing infrastructure and evaluation capabilities.

Mar 25, 2024 | Category: Policy

Anthropic, AWS, and Accenture team up to build trusted solutions for enterprises​

Anthropic announced a collaboration with AWS and Accenture to help enterprises deploy generative AI solutions responsibly, particularly in regulated sectors. Over 1,400 Accenture engineers will be trained as specialists in using Anthropic's models on AWS, providing end-to-end support from concept to production while keeping customer data private and secure.

Mar 20, 2024 | Category: Announcements

Claude 3 models on Vertex AI​

Claude 3 Haiku and Claude 3 Sonnet became generally available on Google Cloud's Vertex AI platform, enabling enterprises to use Anthropic's state-of-the-art models with Google Cloud's infrastructure and tools. This collaboration allows businesses to keep their data within their existing cloud environment while simplifying data governance and reducing operational costs.

Mar 19, 2024 | Category: Announcements

Claude 3 Haiku: our fastest model yet​

Anthropic released Claude 3 Haiku, the fastest and most affordable model in its intelligence class, processing 21K tokens per second and featuring strong vision capabilities. Haiku is three times faster than its peers for most workloads and costs significantly less, enabling businesses to analyze large volumes of documents like quarterly filings or legal cases at half the cost of other models in its performance tier.

Mar 13, 2024 | Category: Announcements

Introducing the next generation of Claude​

Anthropic introduced the Claude 3 model family, setting new industry benchmarks with three models: Haiku, Sonnet, and Opus, each offering different balances of intelligence, speed, and cost. Claude 3 Opus outperforms peers on most evaluation benchmarks and exhibits near-human comprehension, while all models feature sophisticated vision capabilities, 200K context windows, improved accuracy with 2x fewer hallucinations, and significantly fewer unnecessary refusals.

Mar 4, 2024 | Category: Announcements

February 2024​

Prompt engineering for business performance​

Anthropic published guidance on prompt engineering for business performance, emphasizing how effective prompts improve Claude's accuracy, consistency, and cost-effectiveness while reducing hallucination rates. The article shares three key techniques: step-by-step reasoning, few-shot prompting with examples, and prompt chaining for complex tasks.

Feb 29, 2024 | Category: Product

Preparing for global elections in 2024​

Anthropic outlined their preparation for global elections in 2024, implementing policies that prohibit using Claude for political campaigning and lobbying, while conducting targeted red-teaming to test for election-related misuse. The company developed automated systems to detect misinformation and influence operations, and introduced a redirect system for US users asking voting questions to send them to authoritative sources like TurboVote.

Feb 16, 2024 | Category: Policy

December 2023​

Anthropic introduced new Commercial Terms of Service with expanded copyright indemnity protection and launched the beta Messages API for improved developer experience. The updated terms allow customers to retain ownership of outputs and provide legal protection against copyright infringement claims, with Anthropic defending customers and paying for settlements or judgments.

Dec 19, 2023 | Category: Announcements

Long context prompting for Claude 2.1​

Anthropic published insights on long context prompting for Claude 2.1, revealing that while the model has excellent recall across its 200K token context window, it can be reluctant to answer questions based on individual sentences that appear out of place in documents. Adding the simple prompt Here is the most relevant sentence in the context: improved Claude 2.1's performance from 27% to 98% on retrieval tasks.

Dec 6, 2023 | Category: Product

November 2023​

Introducing Claude 2.1​

Anthropic launched Claude 2.1 with significant improvements including an industry-leading 200K token context window (equivalent to 500 pages), a 2x decrease in hallucination rates, and new features like system prompts and beta tool use functionality. The model demonstrates 30% fewer incorrect answers and 3-4x lower rates of mistakenly concluding documents support particular claims.

Nov 21, 2023 | Category: Product

Thoughts on the US Executive Order, G7 Code of Conduct, and Bletchley Park Summit​

Anthropic shared their perspective on three major AI policy developments: the US Executive Order on AI, the G7 International Code of Conduct, and the UK's Bletchley Park AI Safety Summit. The company praised the Executive Order's focus on NIST and the National AI Research Resource, supported the G7 Code of Conduct as setting important baselines for frontier AI companies, and welcomed the Bletchley Declaration's international cooperation.

Nov 5, 2023 | Category: Policy

Dario Amodei’s prepared remarks from the AI Safety Summit on Anthropic’s Responsible Scaling Policy​

Dario Amodei presented Anthropic's Responsible Scaling Policy (RSP) at the UK AI Safety Summit, outlining their AI Safety Levels (ASL) system modeled after biological safety protocols. The RSP includes if-then commitments where dangerous AI capabilities trigger specific safety requirements, with ASL-3 requiring strong security measures and perfect safety performance, and ASL-4 addressing autonomous AI risks.

Nov 1, 2023 | Category: Policy

September 2023​

Claude on Amazon Bedrock now available to every AWS customer​

Claude became generally available on Amazon Bedrock for all AWS customers, providing secure cloud access to foundation models and enabling the development of generative AI applications. The announcement highlighted upcoming Agents for Amazon Bedrock functionality, which allows Claude to orchestrate API calls and perform complex multi-step tasks.

Sep 28, 2023 | Category: Announcements

Expanding access to safer AI with Amazon​

Anthropic announced a strategic partnership with Amazon involving up to $4 billion in investment, making AWS their primary cloud provider and expanding Claude 2 availability on Amazon Bedrock. The collaboration includes access to AWS Trainium and Inferentia chips for model training, secure model customization capabilities for enterprises, and integration across Amazon's businesses while maintaining Anthropic's governance structure through the Long-Term Benefit Trust.

Sep 25, 2023 | Category: Announcements

Prompt engineering for Claude's long context window​

Anthropic published research on prompt engineering techniques for Claude's 100,000 token context window, demonstrating how to maximize information recall from long documents. The study found that using reference quote extraction and providing contextual examples significantly improved Claude's performance on multiple choice questions, with Claude Instant 1.2 showing substantial gains when these techniques were applied to documents containing 70K-95K tokens.

Sep 23, 2023 | Category: Product

Anthropic's Responsible Scaling Policy​

Anthropic released their Responsible Scaling Policy (RSP), establishing AI Safety Levels (ASL) framework modeled after biosafety standards to manage catastrophic risks from increasingly capable AI systems. The policy defines safety requirements for different capability levels, with ASL-2 covering current models like Claude, ASL-3 requiring enhanced security and red-teaming standards, and creates a framework that incentivizes solving safety problems to unlock further AI scaling.

Sep 19, 2023 | Category: Announcements

The Long-Term Benefit Trust​

Anthropic introduced the Long-Term Benefit Trust (LTBT), a governance structure featuring five independent trustees who will gradually gain authority to elect a majority of Anthropic's board within four years. This experimental corporate governance model aims to balance stockholder interests with public benefit by ensuring board accountability to trustees with expertise in AI safety, national security, and public policy, while maintaining the company's mission of developing AI for humanity's long-term benefit.

Sep 19, 2023 | Category: Announcements

Anthropic partners with BCG​

Anthropic partnered with Boston Consulting Group (BCG) to bring Claude to enterprise customers worldwide, focusing on responsible AI deployment across use cases like knowledge management, market research, and business analysis. The collaboration leverages BCG's consulting expertise to help organizations strategically implement AI while maintaining ethical standards, with BCG also using Claude internally for research synthesis and client insights.

Sep 14, 2023 | Category: Announcements

Introducing Claude Pro​

Anthropic launched Claude Pro, a paid subscription plan offering 5x more usage of Claude 2 for $20/month in the US and £18/month in the UK. The service provides priority access during high-traffic periods, early access to new features, and significantly expanded message limits, enabling users to handle more complex tasks like research paper summarization, contract analysis, and extended coding projects.

Sep 7, 2023 | Category: Announcements

August 2023​

Claude 2 on Amazon Bedrock​

Claude 2 became available on Amazon Bedrock, Amazon's fully managed foundation model service, enabling enterprises to build generative AI applications with enhanced security and scalability. Early adopters include LexisNexis using Claude 2 for legal AI capabilities, Lonely Planet for travel planning with 80% cost reduction in itinerary generation, and Ricoh USA for AI-driven operations while maintaining HIPAA and SOC II compliance.

Aug 23, 2023 | Category: Product

SKT Partnership Announcement​

SK Telecom (SKT) became both a commercial partner and strategic investor in Anthropic with an additional $100 million investment, collaborating to develop a fine-tuned large language model optimized for telecommunications applications. The partnership will create a multilingual model supporting Korean, English, Japanese, and Spanish for telco use cases including customer service, marketing, and sales, leveraging SKT's domain expertise to customize Claude for the telecommunications industry.

Aug 15, 2023 | Category: Announcements

Releasing Claude Instant 1.2​

Anthropic released Claude Instant 1.2, an updated version of their faster, lower-cost model that incorporates improvements from Claude 2 while maintaining speed and affordability. The new version shows significant gains in math (86.7% vs 80.9% on GSM8K), coding (58.7% vs 52.8% on Codex), reasoning capabilities, and safety with reduced hallucinations and better resistance to jailbreaks, while generating longer and more structured responses.

Aug 9, 2023 | Category: Announcements

July 2023​

Frontier Threats Red Teaming for AI Safety​

Anthropic conducted frontier threats red teaming research focusing on biological risks, spending 150+ hours with biosecurity experts to evaluate Claude's potential for generating harmful biological information. The study found that current frontier models can produce expert-level knowledge that could accelerate bad actors' efforts, with risks potentially materializing in 2-3 years, but also identified effective mitigations including Constitutional AI training and classifier-based filters that substantially reduce these risks.

Jul 26, 2023 | Category: Announcements

Frontier Model Security​

Anthropic published their approach to frontier model security, recommending two-party control systems and secure software development practices (NIST SSDF and SLSA standards) for protecting advanced AI models. The company advocates treating frontier AI as critical infrastructure requiring enhanced cybersecurity measures, public-private cooperation similar to financial services, and government procurement requirements to ensure industry-wide adoption of robust security practices.

Jul 25, 2023 | Category: Announcements

Claude 2​

Anthropic launched Claude 2, their most advanced model featuring improved performance across coding (71.2% on Codex HumanEval), math (88.0% on GSM8k), and reasoning, with a 100,000 token context window enabling processing of entire books or extensive documentation. The model scored 76.5% on the Bar exam and above the 90th percentile on GRE reading/writing, while being 2x better at harmless responses, and is available via API and the new claude.ai chat interface in the US and UK.

Jul 11, 2023 | Category: Announcements

June 2023​

Charting a Path to AI Accountability​

Anthropic submitted recommendations to the NTIA on AI accountability, proposing a comprehensive framework for evaluating advanced AI systems. The recommendations include funding better evaluations, creating risk-responsive assessments based on model capabilities, establishing pre-registration for large AI training runs, empowering third-party auditors, mandating external red teaming, advancing interpretability research, and enabling industry collaboration on AI safety through antitrust clarity.

Jun 13, 2023 | Category: Announcements

May 2023​

Anthropic Raises $450 Million in Series C Funding to Scale Reliable AI Products​

Anthropic raised $450 million in Series C funding led by Spark Capital with participation from Google, Salesforce Ventures, Sound Ventures, and Zoom Ventures. The funding will support continued development of Claude and AI safety research, including new features like 100K context windows, while growing their product offerings and supporting businesses deploying Claude responsibly.

May 23, 2023 | Category: Announcements

Zoom Partnership and Investment in Anthropic​

Anthropic announced a partnership with Zoom where Claude will be integrated into Zoom's products, starting with the Zoom Contact Center portfolio to improve customer experience and agent performance. Zoom Ventures also made an investment in Anthropic as part of this collaboration, reflecting their shared vision of building customer-centric AI products with trust and security foundations.

May 16, 2023 | Category: Announcements

Introducing 100K Context Windows​

Anthropic expanded Claude's context window from 9K to 100K tokens (approximately 75,000 words), allowing users to submit hundreds of pages of documents for analysis. This breakthrough enables Claude to digest and analyze massive amounts of text in under a minute, supporting use cases like financial statement analysis, legal document review, and codebase modification.

May 11, 2023 | Category: Announcements

Claude’s Constitution​

Anthropic published details about Claude's Constitution, explaining their Constitutional AI approach that uses explicit principles rather than implicit human feedback to guide AI behavior. The constitution draws from sources including the UN Declaration of Human Rights, platform guidelines, and AI safety research, with principles covering helpfulness, harmlessness, honesty, and respect for human values across different cultural perspectives.

May 9, 2023 | Category: Announcements

April 2023​

Partnering with Scale to Bring Generative AI to Enterprises​

Anthropic partnered with Scale to bring Claude to enterprise customers through Scale's robust deployment and management platform. The partnership provides enterprise-grade security, expert prompt engineering, model validation services, and data connectors for proprietary sources, enabling businesses to build and deploy generative AI applications with Claude at scale.

Apr 26, 2023 | Category: Announcements

An AI Policy Tool for Today: Ambitiously Invest in NIST​

Anthropic advocated for ambitious federal investment in NIST to support AI measurement and standards efforts, proposing a $15 million funding increase. The company argues that NIST's century of measurement expertise makes it the natural choice to develop AI evaluation standards, which are essential prerequisites for effective AI regulation and public trust in AI systems.

Apr 20, 2023 | Category: Announcements

March 2023​

Claude, now in Slack​

Anthropic launched the Claude App for Slack in beta, allowing teams to use Claude as a virtual teammate for summarizing threads, answering questions, and various productivity tasks. Claude can be accessed through channel mentions or direct messages, with Anthropic emphasizing that it only sees messages where explicitly mentioned and doesn't use this data for model training.

Mar 30, 2023 | Category: Announcements

Introducing Claude​

Anthropic officially introduced Claude, their next-generation AI assistant based on Constitutional AI research, after testing with partners like Notion, Quora, and DuckDuckGo. Claude offers two versions (Claude and Claude Instant) and is designed to be helpful, honest, and harmless, with early customers reporting it's less likely to produce harmful outputs and easier to steer than alternatives.

Mar 14, 2023 | Category: Announcements

February 2023​

Anthropic Partners with Google Cloud​

Anthropic selected Google Cloud as its cloud provider to support the next phase of scaling and deploying Claude to larger audiences. The partnership enables Anthropic to leverage Google Cloud's GPU and TPU clusters for training, scaling, and deploying AI systems, providing the infrastructure performance needed for their reliable and interpretable AI development goals.

Feb 3, 2023 | Category: Announcements

April 2022​

Anthropic Raises Series B to build steerable, interpretable, robust AI systems​

Anthropic raised $580 million in Series B funding led by Sam Bankman-Fried to build large-scale experimental infrastructure for exploring AI safety properties. The funding supports research into making AI systems more steerable, interpretable, and robust, building on their previous work in language model interpretability and alignment with human preferences.

Apr 29, 2022 | Category: Announcements

May 2021​

Anthropic raises $124 million to build more reliable, general AI systems​

Anthropic raised $124 million in Series A funding led by Jaan Tallinn to support their research roadmap for building reliable and steerable AI systems. The founding team, led by siblings Dario and Daniela Amodei, planned to focus on computationally-intensive research to develop large-scale AI systems that are more interpretable, robust, and better integrated with human feedback.

May 28, 2021 | Category: Announcements

Last updated on