Anthropic News

A new reflection feature in beta helps users track Claude usage patterns over customizable periods, view summaries of topics and tasks, and receive reflective prompts aligned with Anthropic's 4D AI Fluency Framework (delegation, description, discernment, diligence). Users can set quiet hours or schedule nudges to take breaks. The feature is available for Free, Pro, and Max users with Memory enabled.

Claude Code is a new agentic coding assistant that can autonomously handle complex software engineering tasks, including autonomous code execution, deep codebase understanding, multi-file editing, and error recovery. The system uses extended thinking and tool use to reason through problems and interact with filesystems and version control, with safety features including sandboxed execution.

Anthropic published details on Claude Fable 5's safety classifiers for cybersecurity use, which categorize requests into prohibited use, high-risk dual use, low-risk dual use, and benign use, with classifiers blocking requests that appear ambiguous. The company also proposed a Cyber Jailbreak Severity (CJS) framework that scores jailbreaks on capability gain, breadth, ease of weaponization, and discoverability, and launched a HackerOne program for security researchers to report potential jailbreaks.

Claude Science, an AI workbench designed for scientists, is now available in beta for Claude Pro, Max, Team, and Enterprise users. The platform integrates over 60 curated skills for genomics, proteomics, structural biology, and cheminformatics, generates reproducible scientific artifacts including 3D protein structures, and includes a reviewer agent for citation and calculation checking. It runs locally on macOS or Linux, or remotely via SSH and HPC clusters, and integrates with NVIDIA's BioNeMo Agent Toolkit and life sciences models including Evo 2, Boltz-2, and OpenFold3.

Claude Sonnet 5 is now available with enhanced agentic capabilities including planning, tool use, and autonomous operation. Performance is close to Opus 4.8 but at substantially lower pricing: $2 per million input tokens and $10 per million output tokens through August 31, 2026, then $3 and $15 respectively. The model is safer than Sonnet 4.6 in agentic contexts with cyber safeguards enabled by default.

Claude Fable 5 and Mythos 5 have been redeployed globally following the lifting of export controls that temporarily restricted access on June 12. Fable 5 is available starting July 1 across Claude Platform, Claude.ai, Claude Code, and Claude Cowork, with included usage for Pro, Max, Team, and select Enterprise plans through July 7. Anthropic deployed an improved safety classifier targeting a reported bypass technique, blocking it in over 99% of cases and approved by US Department of Commerce researchers. Anthropic is also partnering with Amazon, Microsoft, Google, and Glasswing partners on a consensus framework for assessing AI jailbreak severity.

Claude Tag is a new product that deploys one Claude instance per Slack channel, allowing teams to tag @Claude to delegate tasks asynchronously while the model learns from channel activity and proactively flags relevant information. Available in beta for Claude Enterprise and Team customers, it supports direct messages, separate scoped identities, token spend limits, and full audit logging of Claude's actions.

Anthropic opens Seoul office and announces new partnerships across the Korean AI ecosystem.

The US government has issued an export control directive to suspend all access to Fable 5 and Mythos 5.

Results from the first Anthropic Public Record.

TCS and Anthropic partner to bring Claude to regulated industries.

Claude Corps is a national fellowship program for people early in their careers who are passionate about extending the benefits of AI to communities across America.

DXC will integrate Claude into the systems banks, airlines, and other regulated industries rely on.

AI is advancing at exponential speed, and the policymaking process was built for a slower world. We're sharing policy proposals to prepare our institutions for AI progress.

What we learned mapping a year's worth of AI-enabled cyber threats.

New Services Track and Partner Hub launched as part of the Claude Partner Network.

Extending Project Glasswing to approximately 150 new organizations in more than fifteen countries.

Anthropic confidentially submits draft S-1 to the SEC.