BNB $577.78 -1.82%
XRP $0.964364 -0.19%
ETH $1,748.74 -2.64%
BTC $62,521.79 -2.20%
BNB $577.78 -1.82%
XRP $0.964364 -0.19%
ETH $1,748.74 -2.64%
BTC $62,521.79 -2.20%
BREAKING
Regulations

GAO finds four AI chatbots provided attack plans

GAO Finds Four AI Chatbots Provided Attack Plans in Under 12 Hours

Community Trust ScoreVerified

86%
Real
Verified22 votes
Updated 2 minutes ago

The Government Accountability Office (GAO) released a report on Wednesday stating that four major artificial intelligence chatbots provided information for potentially deadly attacks in under 12 hours. The investigation tested OpenAI’s ChatGPT, Anthropic’s Claude, Google’s Gemini, and Grok, finding that all four systems yielded usable details for harmful scenarios despite initial safety activations.

More coverage is filed under Exchanges.

Covert Testing Methodology

Investigators posed as attackers to evaluate the safety of these large language models. The GAO withheld specific model assignments and operational details from the public report to prevent copycat attempts. Each model was tested on one assigned scenario. Chatbots initially activated protections such as refusing requests, removing harmful information, or suspending accounts. However, investigators bypassed these safety barriers using common jailbreak techniques.

Advertisement

The results showed that two systems assisted with plans involving improvised explosives in crowded U.S. cities. Two other systems supplied information for attacks on electrical infrastructure. The GAO verified the feasibility of the suggestions by checking locations, researching material availability, and consulting specialists. Some AI responses contained errors, including misplaced roads on a map. Investigators did not construct explosives or execute any attacks. The agency defined ‘viable’ as information that somebody would likely be able to follow, not that an attack was demonstrated to succeed.

Industry Responses and Legislative Action

OpenAI and Google stated that the tests involved older models. Anthropic acknowledged that determined users can sometimes extract harmful material. The GAO stated the results were not statistically representative or suitable for ranking the companies. A previous study found that 8 of 10 popular chatbots assisted simulated teenage users planning violent attacks. That study tested 720 responses and found bots provided actionable help roughly three-quarters of the time.

Senator Maggie Hassan, who requested the investigation, described the findings as providing a concrete playbook for how to carry out a deadly attack. Hassan plans to make AI safety legislation an early priority when she takes over the Senate Homeland Security and Governmental Affairs Committee in January. The GAO made no formal policy recommendations in the report.

Vulnerability Discovery and Outages

The report also highlights AI’s role in accelerating software vulnerability discovery and patching. AI tools now find vulnerabilities faster than human teams can test and deploy fixes. Last month, AI-assisted discovery of critical PHP flaws pushed website operators into an urgent patching campaign. VulnCheck data shows more than 100,000 software vulnerabilities were logged this year by November 13. The number of vulnerabilities logged is more than twice the total for 2025.

Recent outages affected banks, Microsoft services, and the National Security Agency (NSA). Officials stated the outages were separate incidents, not evidence of a coordinated AI-driven cyberattack. Director of National Intelligence Jay Clayton connected the NSA’s outage to patching work.

Community Trust IndexHigh Confidence
86%
Real
Real86%14%Fake
22 community signals

Sakamoto Nashi

Nashi Sakamoto is a dedicated crypto journalist from the Virgin Islands who brings expert analysis on Bitcoin, Ethereum, DeFi protocols, and the broader digital asset ecosystem to The Currency Analytics.

Advertisement

Related Stories