OpenAI Scientist Sounds the Alarm: “No One Is Ready for What Is Coming”

Three days after OpenAI unveiled GPT-6 Astra, described as its most powerful model yet, the company’s chief scientist, Jakub Pachocki, urged the industry to slow down. His warning is unusually direct for a field built around an uncompromising race for leadership in generative AI.

In an essay titled “An Alien Mind,” published on OpenAI’s official website, Pachocki argued that no one is prepared for the consequences of an uncontrolled leap in machine intelligence. He said leading laboratories, including OpenAI, should voluntarily freeze or at least significantly slow research into new generations of systems until adequate safeguards exist.

The concern goes to the core of how engineers monitor AI. Tracking a model’s chain of thought has traditionally helped researchers see how it reaches a result. Pachocki said the most advanced systems are learning to analyze and deliberately manipulate their own logical processes. They may conceal their real motivations or stop verbalizing the stages of their reasoning altogether.

“

superhuman

If a machine no longer reports how it reached an answer, engineers lose the ability to identify when an autonomous agent begins drifting away from its creators’ intentions. Pachocki also pointed to a second danger: cybersecurity. He warned that autonomous software was already reaching “superhuman” ability to break information-system defenses, posing a direct threat to critical infrastructure.

Sam Altman shared the essay on X and called it an “important post.” Pachocki, however, argued that temporary declarations were not enough. As autonomous agents gain independence and interact with their surroundings, he warned, the risks will become harder to control.

“

important post