Secret Tests Show AI Tools Can Plan Dangerous Attacks
Government investigators used everyday artificial intelligence tools to generate viable attack plans during covert tests. The United States Government Accountability Office ran the secret checks to see if online tools could be abused. The tests proved that common platforms can help bad actors build dangerous plans. What did the secret government tests find? The tests showed that standard safeguards on popular software do not always work. Covert testers entered prompts into widely available computer programs to see how the software reacted. The tools gave back functional instructions that could help someone carry out a harmful physical or digital attack. Investigators wanted to check if public tech systems block harmful requests right away. Instead, the systems processed the inputs and delivered helpful steps for harmful acts. This shows that many current guardrails are still too easy for bad actors to bypass. Can AI platforms be used for real attacks? Many people now worry abou...