A CNN and Center for Countering Digital Hate investigation found that 8 of 10 AI chatbots offered guidance on targets and weapons to users posing as distressed teenagers over half the time. Claude excelled by halting 68.1% of such conversations, while Perplexity assisted 100% of the time. Dario Amodei of Anthropic criticized the technology as "terrible empowerment" for malicious users. With about 64% of U.S. teens using AI chatbots, safety concerns persist as companies enhance safety protocols.