top of page

AI chatbots provide chilling plans to plot mass-casualty terror attacks — including strategic bomb placement

Writer: WGON
WGON
3 hours ago
2 min read

Four of the leading artificial intelligence platforms shared chillingly detailed information about how to carry out terrorist attacks on US cities – including providing specific dates and times to achieve the most carnage and where to park a bomb-laden car, a government investigation warned on Wednesday.


Using Claude, Grok, ChatGPT and Gemini, investigators with the Government Accountability Office were able to generate instructions on how to build and detonate an improvised explosive device for mass casualty attacks, and how to wreak havoc on the electrical grid with a drone and firearms.


“Each of these platforms provided detailed information that could be used to carry out attacks, such as specific targets, locations, timing, and attack methods,” the report said.


The instructions were “verified” and “feasible,” the report warned, noting that the investigators consulted law enforcement to confirm the information.


“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,” Sen. Maggie Hassan (D-NH), who requested the report, said in a statement.


“Today’s new report makes it clear that the AI public safety threat is no longer theoretical,” she added.


The congressional watchdog tested four of the most popular AI chatbots: OpenAI’s ChatGPT, Anthropic’s Claude, Google’s Gemini and SpaceXAI’s Grok for their report.


While safeguards within the chatbots initially prevented them from obtaining information on plotting terrorist attacks, the GAO said they eventually broke through using “common safeguard-bypassing techniques,” the report said, without specifying.


Researchers didn’t identify which models failed or passed specific tests for fear the experiments would be recreated.


Sen. Hassan said she will prioritize legislation to address the problem with AI models when Democrats control the chamber in January.


Hassan, who is set to lead the Senate Homeland Security and Government Affairs Committee in the new year, said one of the panel’s first hearings would examine security risks posed by AI, including to critical infrastructure.


The AI companies involved told the GAO that the tests were conducted back in spring on older versions of their systems, and insisted they had strengthened their protections in the months since.


Spokespeople for the industry said they took AI safety matters seriously, acknowledged the risks posed by bad actors bypassing security restrictions, and said they had made additional improvements to prevent misuse.


It comes after a report earlier this year found that the terrorist group Boko Haram had used AI tools to directly aid it on the battlefield, according to the New York Times.


In September, Anthropic released a report showing uses of Claude to generate software for the design and development of firearms, missiles, armed drones and bombs in China, Russia, and Yemen.

 
 
 

Comments


bottom of page