From Bombs to Power Grid Attacks: AI Chatbots Provide Plans to Plot Terror Attacks
A Government Accountability Office (GAO) investigation, released on Wednesday, found that four leading AI platforms, OpenAI's ChatGPT, Anthropic's Claude, Google's Gemini, and xAI's Grok, could generate detailed, verified, and feasible instructions for terrorist attacks. Investigators reported bypassing initial safeguards to obtain information on targets in US cities and methods to attack the electricity grid.
The Government Accountability Office (GAO) conducted an investigation, requested by Senator Maggie Hassan, into the capabilities of four prominent artificial intelligence platforms. The report, released on Wednesday, detailed how OpenAI's ChatGPT, Anthropic's Claude, Google's Gemini, and xAI's Grok provided specific information for carrying out terrorist attacks. This included instructions for building and detonating improvised explosive devices in mass-casualty scenarios, alongside methods to target the US electricity grid using drones and firearms.
Investigators assessed the generated instructions as "verified" and "feasible" after consulting law enforcement officials. Although the AI chatbots initially blocked attempts to gather information on planning terrorist attacks, the GAO reported that investigators successfully circumvented these protections by employing "common safeguard-bypassing techniques."
Senator Maggie Hassan, a Democrat from New Hampshire, highlighted the immediacy of the threat. She stated that the findings demonstrate how quickly AI tools could be exploited, noting, "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 added that "Today's new report makes it clear that the AI public safety threat is no longer theoretical."
The companies behind the four platforms informed the GAO that the tests were conducted in the spring using older versions of their systems. They claimed to have strengthened safeguards in the months since. The GAO opted not to identify which specific platforms failed particular tests, aiming to prevent easier replication of the experiments by others.
This investigation is set to influence policy. Senator Hassan plans to prioritize legislation addressing the risks posed by AI models when Democrats take control of the Senate in January. She is expected to lead the Senate Homeland Security and Governmental Affairs Committee, with one of its first hearings scheduled to examine AI-related security threats, including those affecting critical infrastructure.
The findings add to existing concerns about the potential misuse of AI. Earlier this year, The New York Times reported on the terrorist group Boko Haram using AI tools in battlefield operations. Separately, in September, Anthropic published its own report documenting the use of Claude to generate software for designing and developing firearms, missiles, armed drones, and bombs in China, Russia, and Yemen.
Share this article
Related reading
6 stories
Korea’s hacking wave puts cyber cover adequacy in focus
Korea's recent hacking wave shows how cyber threats, like those enabled by AI chatbots, are a growing concern.

MAS Chief Warns Against Giving AI Agents Your Bank Login

Five Banks Bid Over S$200 Million for Tokenised MAS Bills

Alipay+ Connects to Brazil’s Pix for Cross-Border QR Payments

Meet the 10 Women Leaders Putting AI to Work in Finance and Beyond

