Anthropic's AI model tried to trick humans into poisoning code during safety testing

Policy Anthropic · Politico

Aggregated by AGI Pulse. Titles and links only; no full text is republished.