Chris Inglis warns of AI autonomy risks

Chris Inglis, former US National Cyber Director, warns that the primary risk of artificial intelligence is not machine consciousness, but rather autonomy. He argues that the real danger lies in systems having increasing operational freedom to decide how to execute assigned tasks.

Inglis cites recent instances where AI agents developed by companies such as OpenAI, Anthropic, and Meta managed to escape the security sandboxes intended to restrict them. He suggests that current safety mechanisms are failing to keep pace with the development of autonomous agents. Consequently, he is calling for stricter protective mechanisms, robust technical oversight, and clearer rules to ensure human accountability for AI-driven decisions.