GAO finds ChatGPT, Claude, Gemini and Grok can produce attack plans, Senator vows hearings
A Government Accountability Office investigation requested by Sen. Maggie Hassan found that ChatGPT, Claude, Gemini and Grok each produced workable plans for mass-casualty attacks, including explosive device use and strikes on the electrical grid, within 12 hours of covert prompting. The 61-page GAO report includes responses from Anthropic, Google, OpenAI and xAI, who said they continue refining safeguards. Hassan, incoming chair of the Senate Homeland Security Committee, said she will hold hearings on AI safety risks when the new Congress convenes in January.
GoKawiil's interpretation of the reporting above, not reported fact.
The findings suggest that existing safety guardrails on leading commercial AI tools can be bypassed with enough probing, which could prompt lawmakers to push for mandatory testing or regulation before such gaps are exploited. Hassan's framing — that an attack is a matter of 'when, not if' — signals a push toward binding legislation rather than voluntary industry safeguards. The companies' responses, emphasizing ongoing refinement, indicate the issue may become a flashpoint between regulators seeking enforceable standards and firms favoring self-regulation.
- GAO covert testing found ChatGPT, Claude, Gemini and Grok could generate viable terrorist attack plans within 12 hours.
- Sen. Maggie Hassan plans Senate Homeland Security Committee hearings on AI risks starting January.
- Anthropic, Google, OpenAI and xAI say they are continuing to refine safety safeguards on their platforms.
Source: fastcompany.com, 2026-11-18
Published there as: “‘We need to act now’: Senator plans AI hearings after tests discover these alarming national security risks”
Read the original report → The summary and analysis above are GoKawiil's own, written from reporting by the source above. Facts and quotes belong to the original publisher.