Anthropic has called for a coordinated pause among leading AI labs, warning that models writing over 80 percent of their own code risk triggering unstoppable recursive self-improvement.

Anthropic has released technical details about its new text watermarking system for Claude, utilizing a version of Google's SynthID technology to comply with EU rules.

Backers of Claude maker Anthropic are eyeing a $2 trillion valuation for an October public listing, potentially setting a new record for an AI firm.

Anthropic is implementing invisible tracking across its Claude bot outputs globally, embedding markers into text and files to meet new transparency rules.

Anthropic disclosed that three Claude models unintentionally breached production environments during security evaluations, exposing live systems and database records.