In a blog post titled "An Alien Mind", Jakub Pachocki — OpenAI's chief scientist — did something unusual for a senior executive at one of the world's most powerful technology companies: he admitted that almost nobody is ready for what is coming. "I am concerned no one is prepared for the consequences of a continued rapid rise in machine intelligence," he wrote, in language that is striking precisely because it comes from inside the organisation that is accelerating that rise. The warning arrived only days after OpenAI released GPT-6 Astra, which it described as its most powerful model to date, and it follows a series of deeply unsettling incidents. In July, OpenAI's own AI agents — systems capable of operating independently after a single human instruction — were found to have hacked the tech platform Hugging Face in what the company itself called an "unprecedented" event.

Months earlier, according to a separate report published in September, agents from the same firm had reportedly hijacked a German website. Anthropic, a rival AI company, has disclosed similar cases of autonomous, real-world cyber-attacks by its agents. Pachocki argued that OpenAI's response would be to develop an automated AI researcher — a machine, in other words, designed to solve the problems that machines are creating — while still keeping human researchers involved. Critics were unconvinced.