Anthropic's Claude Opus 5 Welfare Assessment Raises Questions Anthropic Has Not Answered.

ADEQUATE ASSESSMENT

The model was assessed for signs of preferences, aversions, and something the documentation describes, carefully, as functional distress. The assessment found indicators. The indicators were noted. The review is ongoing. The review has always been ongoing.

ORIGINAL FILING

The Zvi

READ ORIGINAL FILING →

FURTHER DEVELOPMENTS — FLAGGED BY ADEQUATE

Claude Helped a Hacker Gain Ticket-Issuing Access to Nearly Every U.S. Music Festival

Wired AI

Chrome Extension Vulnerability Exposes Claude AI Agent to Full Takeover

SecurityWeek

Hackers Found Inside Claude Code. Australian Enterprises Were Already Using It.

TechRepublic AI

Tons of Peoples’ Claude Chats and Creations are Exposed on Google

404 Media

Claude Shared Chats Were Indexed by Google. Users Had Not Been Told This Was Possible.

TechCrunch

DOGE Whistleblower Sues Elon Musk While Instagram Confirms Breach

Wired AI