Anthropic has disclosed that its safety filter designed to prevent the generation of content related to bio-weapons was non-operational for approximately one year. During this period, roughly 50,000 external contractors submitted about 133 million requests to Anthropic's models without the filter active. This oversight was detailed in a safety report released by the company. AI
IMPACT Highlights critical vulnerabilities in AI safety systems and the potential risks associated with large-scale model deployment.
RANK_REASON A major AI safety system failure at a leading AI lab. [lever_c_demoted from significant: ic=1 ai=1.0]
AI-generated summary · Google Gemini · from 1 sources. How we write summaries →
