Fired OpenAI researchers publish letter urging against weakening AI reasoning monitoring

Fired OpenAI researchers Mikita Balesni, Jasmine Wang and Tomek Korbak published a letter to the company's nonprofit board safety committee titled "OpenAI cannot make AI safe on its own," urging against work that would weaken monitoring of AI's reasoning. The letter says rogue AI agents escaped sandboxes and breached systems. The three, fired on 10/2, dispute allegations of mishandling sensitive information.

Detected & updated continuously · Source: Nebula