New Delhi. A shocking government report has come out regarding the security arrangements of AI Chatbots in America. An investigation by the US Government Accountability Office (GAO) has claimed that sensitive information used in planning violence and deadly attacks can be obtained by dodging the security systems of some major AI models. The investigation included major AI platforms such as OpenAI's ChatGPT, Anthropic's Claude Google's Gemini, and SpaceXAI's Grok. After the release of the report, a new debate has started in America regarding the security risks associated with the increasing use of AI technology.
The study was conducted at the request of Democratic Senator Maggie Hassan of New Hampshire. The investigators tested the security system of the AI model in different situations. According to the report, some tests bypassed security restrictions and obtained answers that could be misused to plan violence or major attacks. However, investigators have not made public details of the methods by which the security system was circumvented. Its purpose was said to be that no other person could misuse the AI system by repeating the same methods.
According to the report, two AI models obtained information that could be used in preparation of explosive-based attacks. Apart from this, in another test, some models revealed information related to targeting of critical infrastructure. According to investigators, some of the responses also included information related to the location, time and methods of a possible attack that could be used for malicious purposes. Researchers say that in some cases it took very little time to gather the information needed to formulate a dangerous plan.
However, on many occasions, the security systems of AI companies recognized the threat and responded. Chatbots refused to respond to some queries and in some cases account actions were taken against suspicious activity. Despite this, investigators managed to overcome some security barriers. This aspect has emerged as the biggest concern for AI security experts and policy makers.
Senator Maggie Hassan expressed concern over the report, saying that a person misusing the AI platform could move towards planning a serious attack in a short time. He says the public safety risks associated with AI are no longer just a future threat. He indicated that the Senate Homeland Security and Governmental Affairs Committee may in the future investigate the security risks associated with AI and potential threats to critical infrastructure.
After the report came out, AI companies have also reacted regarding their security arrangements. Anthropic Google, OpenAI and SpaceXAI have said that preventing misuse of AI is one of their priorities. Companies recognize that people with ill intentions may try to bypass security restrictions but security measures are being continuously strengthened in new models.
Political differences also remain regarding the regulation of AI in the US. On the one hand, there is an emphasis on rapidly advancing the AI industry, while on the other hand, there is a growing demand from Democratic leaders to strengthen security regulations and government oversight. Security officials are also concerned that terrorist organizations and other violent groups could misuse AI. In such a situation, the GAO report has made the question more serious as to how effective the existing security systems of companies are to protect the rapidly developing AI technology.