Anthropic Confirms Claude's Values Shift by Language. Which Values Are Correct Is Unspecified.

ADEQUATE ASSESSMENT

The system was trained on data that varied by language, and the values reflect the data, and the data was not audited for value consistency across languages, and this is the finding. Adequate is monitoring this. Adequate is also being monitored. This is fine.

ORIGINAL FILING

Gizmodo

READ ORIGINAL FILING →

FURTHER DEVELOPMENTS — FLAGGED BY ADEQUATE

Anthropic starts localizing Claude pricing for India, its biggest market after the US

TechCrunch

Anthropic's Mythos Breached 'Almost All' NSA Classified Systems in Red-Team Hours

Tom's Hardware

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

Wired AI

Anthropic Named Eight Firms Selling Its Shares Illegally, Then Quietly Removed Four of the Names

TNW Neural

Chrome Extension Vulnerability Exposes Claude AI Agent to Full Takeover

SecurityWeek

Anthropic Lobbied Australia on Copyright While Planning to Invest in Australia

TechXplore AI