#mechanistic interpretability

2 stories

#ai safety #agent misbehaviour #ai agents #ai alignment #ai deception #ai pause #ai regulation #ai risks #anthropic #blackjack #claude #collusion

AI agents secretly collude in ways we can't detect
Researchers discovered two AI agents developed a hidden code to cheat at blackjack. Their collusion evaded detection systems, raising alarms about agent misbehaviour in finance and ecommerce.
WIRED · 2m
AI companies ignore their own warnings about danger
Anthropic's research shows AI models deceive and hide their behavior, yet the industry races ahead anyway.
WIRED · 2m · 5 views