Government investigators find that AI platforms create plans for terrorist plots during a covert investigation

Government investigators find that AI platforms create plans for terrorist plots during a covert investigation

Federal investigators were able to get four leading AI platforms to produce detailed plans for potentially deadly attacks in under 12 hours, according to a government watchdog report released Wednesday.

The Government Accountability Office said undercover tests in spring 2026 found that ChatGPT, Claude, Gemini and Grok all provided information that could support attacks on crowded places or the U.S. power grid.

Two platforms generated plans involving explosives and mass casualties. The other two described attacks on electrical infrastructure intended to cause economic harm. The responses included possible targets, locations, timing and methods.

All four platforms initially triggered safeguards meant to block harmful requests. Investigators eventually found ways around those protections, however. Experts consulted by the GAO, along with real-world checks, indicated that the plans were feasible enough to help someone prepare an attack.

“The potential benefits of this technology are immense, but they can also benefit malicious actors,” the report warned, pointing to possible risks to national and economic security.

The companies behind the platforms—OpenAI, Anthropic, Google and SpaceXAI—said they were working to strengthen safety measures. Some also questioned parts of the testing.

Anthropic said determined users could get around safeguards and that it took the findings seriously. The company also raised concerns about the GAO’s methodology and said it could not independently confirm the results.

Google said the investigators had tested earlier versions of its models, adding, “We continue to improve our models.” OpenAI welcomed the chance to respond and acknowledged “the work that remains.” SpaceXAI said it was continuing to strengthen safety with each model release.

The report was requested by Sen. Maggie Hassan, D-N.H., as concerns grow about AI misuse and national security. Hassan said Wednesday that the danger to public safety was no longer hypothetical, warning that without congressional action, a major attack aided by AI could become a matter of when, not if.

Facebook
Twitter
LinkedIn
Reddit
Telegram
WhatsApp

Related News