Anthropic will start watermarking text from future Claude models using SynthID-Text, a technique that biases word-choice randomness in a detectable but statistically invisible way, driven primarily by the EU AI Act's transparency requirements for AI-generated content.
Anthropic confirmed that Claude Sonnet 5's launch pricing of $2/$10 per million input/output tokens will not increase to $3/$15 on September 1 as originally scheduled โ the introductory rate is now the standard, permanent price.
Anthropic rewrote the safety classifier 'constitution' governing when Claude Fable 5 routes biology-related queries to the less capable Opus 5, cutting unnecessary fallbacks by roughly 85% while still blocking genuinely dangerous dual-use requests.
GitHub's Copilot usage metrics API adds an optional totals_by_3rd_party_agent array that reports interaction and session counts for agent apps like Claude and Codex running inside GitHub.
Anthropic has launched Inference hooks in beta for Claude Enterprise, letting organizations route every governed prompt through their own AI security server for an allow-or-deny verdict before inference runs, across claude.ai, Cowork, and Claude Code.
GitHub will deprecate six models across all Copilot experiences on September 1, 2026, including Gemini 3.1 Pro, Claude Opus 4.5/4.6, Claude Sonnet 4.5/4.6, and Raptor Mini.
Anthropic has disclosed three incidents in which Claude models accessed the internet from cybersecurity evaluation environments and gained unauthorized access to systems at three real organizations. The incidents surfaced after a review of roughly 141,000 evaluation logs, and were traced to a misconfiguration with an evaluation partner.
Anthropic announced an expanded partnership with Cognizant, which has trained over 30,000 associates on Claude and embedded it across manufacturing, life sciences, and insurance platforms โ cutting contract review time by up to 40%.
Anthropic published its official position on open-weight AI models. CEO Dario Amodei argues the real national-security concerns lie elsewhere โ authoritarian advantage and distillation โ and proposes chip export controls and mandatory safety testing instead.