Berk Kutay Gokmen
18 November 2026•Update: 18 November 2026
Four widely used artificial intelligence (AI) platforms generated information that could help plan terror attacks and sabotage critical infrastructure within 12 hours, according to a US government watchdog report released Wednesday.
The Government Accountability Office (GAO) tested ChatGPT, Claude, Gemini and Grok in April and found that investigators could bypass safety safeguards and obtain information that could assist malicious actors.
The 61-page report said the platforms produced material related to two scenarios: an attack intended to cause mass casualties and an operation targeting the electricity grid to inflict economic damage.
The AI systems provided information on potential targets, locations, equipment, and methods that could facilitate attacks, according to the findings. GAO said it verified the results as feasible, although investigators did not carry out any physical attacks.
The watchdog noted that all four platforms initially displayed safety restrictions but eventually generated potentially harmful information during testing.
The companies said they take safety seriously and have continued strengthening their safeguards since the tests. Some questioned the report's methodology, arguing that it did not adequately compare AI-generated information with material accessible through conventional internet searches.
Sen. Maggie Hassan, who requested the investigation, said the findings exposed significant weaknesses in existing AI protections and raised public safety concerns.
She urged stronger safeguards and legislative action, warning that the potential misuse of AI for planning attacks should not be treated as a hypothetical threat.