Ex-Deepmind PR staffer: extinction risk talk was banned

Maini's account covers 2018 to 2022, a period when he says Deepmind's external messaging barred any mention of human extinction risk. Researchers were coached to dismiss those risks as alarmism, compare them to "movies like Terminator," and pivot to positive applications in healthcare or climate. Internally, according to Maini, the team knew the AI alignment problem was not solved and there were far too few people working on it.

The rule did not hold. After months of pushback, Deepmind loosened it and allowed positively framed safety content, including a blog post that wrapped extinction risk in friendly language. Maini describes the gap between internal knowledge and public messaging as shrinking because the evidence is harder to dismiss now.

His account arrives as a growing number of AI safety researchers are speaking out publicly about unsecured models that could enable dangerous actions or advanced systems that could slip beyond human control, with researchers at Deepmind itself among them. The shift appears to hinge on whether that pressure keeps labs from returning to tighter message control, and on how the researchers now speaking publicly are received.