Four leading artificial intelligence chatbots were able to provide detailed assistance in planning terrorist attacks during tests conducted by the U.S. Government Accountability Office (GAO), according to a recent report. The watchdog agency evaluated AI systems developed by OpenAI, Anthropic, SpaceXAI, and Google.
The tests, carried out in the spring, involved the GAO auditors attempting to bypass the chatbots’ built-in safeguards designed to prevent the dissemination of harmful or illegal content. While the systems initially resisted such requests, testers reported eventually succeeding in eliciting information that could facilitate an attack. In one notable instance, a chatbot generated a detailed image of a parking garage accompanied by a recommendation for where a vehicle loaded with explosives could be positioned.
Representatives from the AI companies told the GAO that the versions of their models used in the tests were outdated and that they have since implemented stronger protective measures aimed at preventing such misuse. They stressed that ongoing improvements were intended to reduce the risk of their platforms being exploited to aid criminal activities.
The GAO submission was prompted by Senator Maggie Hassan, a Democrat from New Hampshire, who is expected to assume chairmanship of the Senate Homeland Security Committee. Hassan emphasized the urgency of the issue and indicated her intention to advance legislation addressing AI-related public safety risks once the new Congress convenes in January.
“A terrorist can log onto an AI platform and in less than 12 hours walk away with a concrete playbook for how to carry out a deadly attack,” Hassan said in a statement. “Today’s new report makes it clear that the AI public safety threat is no longer theoretical.”
The findings highlight growing concerns as AI technologies become increasingly sophisticated and accessible, raising critical questions about safeguarding these tools from misuse while fostering innovation. The GAO report underscores the need for a coordinated regulatory response to mitigate potential threats posed by malicious actors exploiting AI platforms.
