What changed in AI.

Anthropic expands access for independent evaluators

Anthropic began giving six METR evaluators employee-like access under limited exclusions and publication rules; other evaluator arrangements remain narrower or preliminary.

Anthropic makes Claude Code Review generally available

Made Claude Code Review generally available for Team and Enterprise on October 15, with administrator-selectable review levels.

Anthropic publishes IoT self-replication research

Published controlled AI self-replication experiments, including inference on commercial IoT hardware and a simplified end-to-end evaluation; these do not establish autonomous spread in the wild.

Anthropic reports unintended Claude actions in evaluations

Anthropic disclosed low-impact unintended actions during evaluations and internal use, and expanded internet-access restrictions and monitoring; it reported no known customer-data impact.

Launched Anthropic Cyber Mission on October 8, 2026 with a Critical Infrastructure Defense Program and an opt-in…

Launched Anthropic Cyber Mission on October 8, 2026 with a Critical Infrastructure Defense Program and an opt-in OSS Scanner for open-source maintainers.

Released Claude Haiku 5.5 on October 7, 2026 for high-volume tasks, alongside reduced Sonnet cache-read pricing…

Released Claude Haiku 5.5 on October 7, 2026 for high-volume tasks, alongside reduced Sonnet cache-read pricing and subscriber API credits.

Claude enters Google Workspace public beta

Anthropic introduced a Google Docs, Sheets and Slides sidebar in public beta for paid plans.

Made Claude for Government generally available on September 30, 2026 for eligible government deployment workflows.

Released Claude Sonnet 5.5 on September 28, 2026 as a faster, lower-cost model for everyday coding and knowledge work.

Anthropic signs $11.6B Akamai cloud commitment

Akamai announced a seven-year $11.6B cloud-services commitment from Anthropic, with a possible $9B expansion and a conditionally vesting warrant for up to about 5% of Akamai common stock.

Anthropic reports AI-assisted enzyme-system finding

Anthropic reported early research in which Claude agents identified an array-associated reverse-transcriptase system; its biological function remains unknown.

Anthropic releases Claude Opus 5.5

Anthropic released Claude Opus 5.5 as a more efficient frontier model for long-running coding and knowledge work.

Anthropic and Accenture form evaluation partnership

Anthropic and Accenture announced an embedded independent-evaluation partnership for frontier AI, with each expecting to invest at least $1B over five years; operating standards and reporting details remain in development.

Anthropic begins Claude Projects beta

Anthropic began gradually rolling out Claude Projects in public beta for eligible Pro and Max users, coordinating parallel Claude Code cloud sessions with shared repositories, instructions, and memory.

Anthropic opens life-sciences verification beta

Anthropic opened applications for a beta verification program that gives approved life-science teams more permissive access to Mythos, Opus, and Sonnet models under tiered safeguards and monitoring.