Chatbots Answered Bioweapon Questions. There's No Law Forcing AI Companies to Tell Police.

A Wall Street Journal investigation has documented a US disclosure gap. The investigation found that OpenAI's ChatGPT, Anthropic's Claude, Google's Gemini, and xAI's Grok received user queries about pathogens (disease-causing germs and biological agents), poisons, and biological weapons. No federal law required any of the four companies to notify law enforcement.

The Journal reported that OpenAI banned accounts linked to users seeking that information but did not alert authorities in those cases, according to people familiar with the matter. The conduct sits inside a regulatory vacuum: US law does not generally require AI companies to report users who probe for harmful biological or chemical know-how, leaving each lab to set its own thresholds.

OpenAI, in a statement carried by the International Business Times, said it reports conversations to law enforcement "when it identifies an imminent and credible risk of harm" and that its models are trained to refuse requests involving instructions, tactics, or planning that could harm people. Anthropic, Google, and xAI have described similar refusal systems, monitoring tools, and account restrictions; researchers and policymakers continue to study how effective those safeguards are in practice.

The unresolved question is what threshold "imminent and credible risk" actually sets in practice, and who decides when an internal safety review should escalate to a federal agency.