Sam Altman says AI is finding ‘Insane’ hardware-level bugs we never expected

Altman’s recent remarks come after the PHP security vulnerabilities that hit websites across the world OpenAI CEO Sam Altman has raised concerns over growing capabilities of artificial intelligence in finding security vulnerabilities beyond our expectations. In a recent interview, Altman explained that it is becoming hard to restrict or confine AI to the sandboxes as they just work for limited time. Once the model starts evolving its cutting-edge abilities, you can’t even find the traces of how and when they escaped. While ongoing monitoring buys time, Altman notes that true safety requires AI alignment: getting the model to internalize human values. With the risks of misalignment at an all-time high, he stresses that alignment research is still in its early stages and far from resolved.