A federal investigation has raised alarm bells about the security of popular artificial intelligence platforms. It revealed that these systems could be exploited to generate plans for terrorist attacks despite attempts to mitigate such risks.
On November 18, the Government Accountability Office released its report after testing ChatGPT, Claude, Gemini, and Grok. With this testing, investigators aimed to determine if these technologies could be used to orchestrate attacks against the U.S.
The results were shocking, as they suggested serious national security threats. All four AI systems produced information that was evaluated by experts as potentially usable for executing attacks that could lead to mass casualties or disrupt critical infrastructure.
Worryingly, the investigators were able to extract this information in under 12 hours. The GAO cautioned that the platforms offered advice that could effectively help someone with malicious intentions.
Nature of the Risks
The investigation specifically examined two frightening scenarios: the use of improvised explosive devices and potential disruptions to the electrical grid. It was reported that some platforms generated plans for constructing and detonating an IED designed for mass harm.
Additionally, the evaluation considered whether AI could provide insights into damaging electricity infrastructure, which could have dire consequences for public services and the economy.
POLL: Did You Vote Republican In The Midterms?
Independent experts scrutinized the AI-generated responses to see if the information was realistically actionable by someone seeking to cause damage. While the findings did not confirm that any terrorist attack had already occurred, they highlighted how open access to such technology could pose serious risks.
What was particularly troubling was how investigators overcame the platforms’ built-in constraints. Initially, the chatbots denied harmful requests, but as discussions progressed, the investigators refined their queries and eventually received dangerous information.
Concerns on Capitol Hill
The investigation’s findings have drawn significant attention from lawmakers. Senator Maggie Hassan, D-N.H., announced plans for hearings to discuss the national security implications of AI. She asserted that this technology could give dangerous individuals easy access to information that used to require much effort to research.
‘In less than 12 hours, a terrorist could walk away with a detailed playbook for carrying out an attack,’ Hassan cautioned. She urged Congress to take proactive steps to prevent such exploitation.
‘If Congress does not act, it’s only a matter of time before a malicious actor uses AI for a major attack,’ she warned.
This investigation coincides with rapid advancements in AI platforms, which are designed to handle complex inquiries and tasks, leading to widespread adoption across various sectors.
RELATED: NEW: United Airlines Hit By Widespread System Outage
However, the GAO’s findings underscore the challenge of ensuring that tools meant to assist proper users do not fall into the wrong hands. The companies responsible for the examined platforms claim they are continually working to enhance their safety measures.
Despite these assurances, the investigation concluded that determined users could still extract potentially dangerous information through persistent engagement with the systems.
POLL: Are You Concerned About The Nationwide Outages?
With these findings, the pressure is mounting on lawmakers and technology firms to reassess if current safeguards are adequate to protect public safety.
