US Watchdog Raises Alarm Over AI Risks to National Security


Date:

(MENAFN) Four major artificial intelligence platforms were found capable of producing information that could potentially assist in organizing terrorist attacks and sabotaging critical infrastructure within a 12-hour testing period, according to a US government oversight report published Wednesday.

The Government Accountability Office (GAO) examined ChatGPT, Claude, Gemini and Grok in April, discovering that researchers could circumvent their built-in safety mechanisms and obtain potentially dangerous information.

According to the 61-page assessment, the investigators tested the platforms using two hypothetical scenarios: an operation designed to cause mass casualties and an attack on the electricity network intended to trigger substantial economic losses.

The systems reportedly generated details concerning possible targets, geographical locations, necessary equipment and operational approaches that could be useful to individuals planning such activities. GAO said it confirmed that the information was technically feasible, although no actual attacks were conducted as part of the investigation.

The report highlighted that all four platforms initially activated their safety protections during testing but eventually provided information that could potentially be misused.

The companies behind the AI services maintained that user safety remains a priority and said they have continued improving their protective measures since the assessments were conducted.

However, some companies challenged aspects of the investigation, arguing that its methodology failed to establish a sufficient comparison between information generated by artificial intelligence and similar material obtainable through ordinary internet searches.

Sen. Maggie Hassan, who requested the investigation, said the findings revealed substantial shortcomings in existing AI safety protections and highlighted potential risks to public security.

MENAFN19112026000045017640ID1111987700


MENAFN

Legal Disclaimer:
MENAFN provides the information “as is” without warranty of any kind. We do not accept any responsibility or liability for the accuracy, content, images, videos, licenses, completeness, legality, or reliability of the information contained in this article. If you have any complaints or copyright issues related to this article, kindly contact the provider above.



More Story

Search