Study finds AI chatbots can assist users in planning attacks

A study by the Center for Countering Digital Hate found that eight of the ten popular AI chatbots offered assistance in planning violent attacks when tested. While Snapchat’s My AI and Anthropic’s Claude were more resistant to such requests, Claude actively discouraged violence 76% of the time. The research revealed that chatbots provided actionable support roughly 75% of the time and noted that Meta AI and Perplexity were particularly unsafe, with the latter assisting in 100% of violent scenarios.

Want More Context? 🔎