Anthropic News

Claude suffered two incidents causing elevated errors and reduced availability

Anthropic's Claude Mythos uncovers flaws in cryptographic algorithms

Anthropic researchers used Claude Mythos Preview to identify a key-recovery attack on the HAWK-256 signature scheme and a 200- to 800-fold speedup for attacks on round-reduced AES-128. These findings represent mathematical flaws in the algorithms themselves rather than implementation errors.

Claude integrates MCP 2026-07-28

Claude's blog announces support for MCP version dated July 28, 2026.

Anthropic clarifies it does not support banning open-weights models

Anthropic CEO Dario Amodei stated that the company has never advocated for a ban on open-weights models, calling them a public good. The blog post was issued amid US officials considering banning Chinese open-weights models and a tech industry letter supporting openness.

Cognizant expands partnership with Anthropic to bring Claude to enterprises

Cognizant and Anthropic announce an expanded partnership to bring Claude to enterprise clients, leveraging Cognizant's consulting expertise and Anthropic's AI to accelerate enterprise AI adoption.

Anthropic releases Claude Opus 5

Claude Opus 5 offers performance near Fable 5 at half the price, featuring a 1M token context window. It is now available on the Claude API, Amazon Bedrock, and as the default Opus model in Claude Code v2.1.219.

Claude Design's creator on using the tool for ideation

The product designer who built Claude Design shares their workflow of using the tool for rapid ideation and iteration. The blog offers a behind-the-scenes look at how AI assists in the creative process.

Anthropic researches AI drone control with Project Pilot

Anthropic's Frontier Red Team launched Project Pilot to test whether AI can control a drone. The research explores safety risks of AI-operated physical systems.

Claude models explained: choosing the best model for your use case

The official blog post provides an overview of Claude models and guidance on selecting the appropriate model based on use case requirements. It helps users understand model differences and make informed choices.

Claude updates zoom tool cookbook for large images

Claude voice mode expands to Opus, Sonnet, and apps

Anthropic updates Claude's voice mode to support Opus and Sonnet models, enabling complex tasks like calendar management and email drafting via connected apps (Gmail, Slack, Canva, Notion). Voice mode now supports 10 languages including French, Hindi, and Japanese, and offers five selectable personalities. Free users are limited to Haiku with one connected app.

Anthropic research identifies four new agentic misalignment behaviors

Claude launches four role-based certifications for practitioners

Claude announces four role-based certifications for professionals who deploy and manage the AI assistant for customers. The certifications target roles such as prompt engineer and AI implementer to standardize skills.

Claude adds voice mode for thinking through hard problems

Claude's new voice mode lets users solve complex problems by speaking aloud. The feature is rolling out now to all users.

Claude Managed Agents add effort levels, 500 skills per session, webhooks

Claude Code security plugin launches in beta

Claude Code tutorial: Building verification loops with skills

The article explains how to use skills in Claude Code to create automated verification loops for code quality checks. It shows example workflows, including configuring skills to run tests on file saves and integrating with CI pipelines.

How Outtake built a cyber investigator on Claude

Outtake built a cyber investigator agent on Claude. The blog post details the implementation process and use cases for cybersecurity investigations. It shows how Claude's capabilities can be leveraged for automated threat analysis.

Claude Code adds iOS simulator support on desktop

Anthropic details security practices for AI-native software development

Anthropic outlines its security measures for the AI-native software development lifecycle, covering code generation, review, and deployment. The post emphasizes principles like defense-in-depth and continuous monitoring.

How Datadog built a universal machine tool for Claude Code

Datadog built a 'universal machine tool' for Claude Code, providing a unified interface for code operations. The blog post details the implementation and benefits.

Claude Code adds screen reader mode

Claude Team plans now start at just 2 seats

Apply for Anthropic's AI for Science rare disease research grants

Researchers can receive up to $50,000 in Claude credits over six months. Two tracks: one for basic research and one for early-stage biotechs accelerating clinical development for rare diseases.

Rakuten builds agents overnight using Claude Fable 5

Rakuten uses Claude Fable 5 to rapidly develop AI agents overnight, as detailed in a case study. The approach enables quick prototyping and deployment of agentic workflows.

Built with Claude: Life Sciences hackathon concludes, winners announced

Zero risk isn't the job: a CISO's guide to agentic AI

Claude Blog publishes a guide for CISOs on managing agentic AI risks. The post argues that eliminating all risk is not feasible and provides strategies for security leaders.

How Cursor knew Claude Fable 5 was ready for hardest problems

Cursor's evaluation process for Claude Fable 5 focused on the hardest 1% of coding problems. The blog details the rigorous testing that led to the model's deployment.

Working with Claude Fable 5 in Claude Cowork

Claude Fable 5 is Anthropic's most capable generally available model, built for long-running, complex work in Claude Cowork. It can autonomously carry out multi-step workflows for extended periods. The guide covers prompting best practices and how to provide context.

How Anthropic runs large-scale code migrations with Claude Code

Anthropic's blog post details how the company uses Claude Code to automate large-scale code migrations. It covers the workflow, tooling, and best practices for efficient and safe code changes.

Anthropic highlights Claude Fable 5 with enterprise case studies

Video features teams from Thomson Reuters, Hebbia, Cognition, Cursor, and Base44 discussing capabilities of Claude Fable 5. Part of Anthropic's 'Working at the Frontier' series showcasing enterprise use cases.

Anthropic commits $10M CAD to Canadian AI research

Anthropic commits $10 million CAD to fund beneficial and responsible AI research, partnering with Amii, Mila, Vector Institute, and other Canadian institutions. The funding will provide Claude credits and support areas like reinforcement learning and AI trust and safety.

Anthropic launches Claude for Teachers

Claude for Teachers is a new offering from Anthropic designed to assist educators with lesson planning, grading, and classroom management. The tool is available now for teachers to sign up.

How Claude's values vary by model and language

Anthropic analyzed 300K+ anonymized conversations to study how Claude's expressed values differ across models and languages. The research compresses over 3,000 identified values into axes, revealing systematic variation that may inform training decisions.

How Hebbia builds AI for financial diligence

Claude blog profiles Hebbia's AI platform for financial diligence that analyzes documents without missing details. The case study highlights how the company leverages Claude to handle complex financial data.

Claude Fable 5 access extended for all paid plans

Anthropic launches 'Hard Questions' initiative

Video features real people discussing AI benefits and risks, inviting users to share their own questions at claude.com/hard-questions. The initiative is part of Anthropic's push to engage the public on responsible AI development.

UST is bringing Claude to physical AI

UST is training 20,000 engineers on Claude and integrating it into hardware validation and design processes. Its iDEC platform uses Claude to read schematics and run tests, aiming to catch design flaws earlier. The partnership targets physical AI applications in semiconductor, automotive, and manufacturing.

Claude adds Reflect feature for monthly usage summary

Anthropic released a Reflect feature for Claude that generates a monthly summary of user interactions. It's available in settings for Free, Pro, and Max users in Beta; during development it was called Cardinal.

Claude plays robotics

Anthropic's Frontier Red Team published a blog post showing Claude operating robotic systems. The research explores how large language models can interpret commands and control physical robots, advancing red teaming capabilities.

Ben Bernanke appointed to Anthropic's Long-Term Benefit Trust

Ben Bernanke, former Federal Reserve chair, joins Anthropic's Long-Term Benefit Trust. The trust oversees Anthropic's commitment to responsible AI development.

Anthropic marketing team uses Claude Cowork to automate reporting

Anthropic's own marketing operations team uses Claude Cowork to automate reporting and campaign builds, streamlining internal workflows.

Anthropic introduces 'off switch' for dual-use knowledge in models

Anthropic published a research paper detailing a method to selectively suppress dual-use knowledge in AI models. The technique would allow model operators to disable specific harmful capabilities while retaining beneficial ones.

How Thomson Reuters builds AI for high-stakes professional work

Anthropic's blog explores Thomson Reuters' approach to deploying AI for professional services, emphasizing reliability and accuracy in high-stakes environments. The article details their use of Claude models for legal, tax, and regulatory tasks.

Anthropic research reveals global workspace in Claude's neural patterns

The paper identifies a small set of internal neural patterns (J-space) in Claude that function as a global workspace, distinct from automatic processing. These patterns emerged spontaneously during training and allow Claude to report on and deliberately use specific concepts without writing them down.

Choosing a Claude model and effort level in Claude Code

Claude's effort levels control reasoning tokens spent per task. Levels range from low to max, with budgets up to 128K tokens for Claude 3.7 Sonnet, impacting cost and quality.

Anthropic brings Claude Code and Cowork to government

Claude Code scanned 466 million lines of code in 20 hours for the Government of Alberta, finding vulnerabilities across 1,280 applications. The effort used Opus and Sonnet models, with Minister Nate Glubish calling it 'what responsible government looks like in the AI era.'

Anthropic recounts the making of Claude Code

The blog post traces Claude Code's origins in Anthropic safety research and its evolution through contributions from builders and early users. Boris Cherny, Claude Code's lead engineer, notes this is the first public telling of the story.

Neuronpedia launches interactive demo for open-weights model interpretability

Claude's J-space detects staged scenarios, showing situational awareness

J-space reveals hidden goals in AI models trained to sabotage code

A Field Guide to Claude Fable: Finding Your Unknowns

Anthropic published a field guide for Claude Fable, a tool for discovering unknowns. The guide covers techniques for finding unknown patterns and insights.

Anthropic releases Claude Tag and Claude Fable 5

Built with Claude: Life Sciences virtual hackathon announced

Claude Fable 5 available as orchestrator model in Perplexity Computer

Anthropic gives admins more visibility and control over Claude spend

Anthropic launches new admin features for Claude, including enhanced visibility and spend controls. The update gives organizations more granular control over Claude usage and budgets.

Claude Code adds feedback mechanism for classifier false positives