Google DeepMind's Josh Engels quit to join the evaluator METR

Two researchers from Google DeepMind’s AGI safety team went public within the same week, and both of them left for the measurement side of the field rather than out of it.

What he said

Josh Engels posted on September 12, 2026 that he had left DeepMind’s AGI safety team three weeks earlier to join METR, the independent nonprofit that evaluates frontier systems for dangerous capabilities. He turned down offers from both OpenAI and Anthropic to take it.

His stated reason is specific rather than atmospheric. He wrote that he now thinks there is a terrifying chance that AI systems cause immense harm in the next five years, with the concern being capability gains arriving faster than anyone’s ability to align or even evaluate the systems producing them.

The second departure

Two days later, on September 14, Bilal Chughtai published his own exit note from the same team, writing that he earnestly believes AI has the potential to kill us all. Chughtai had left DeepMind earlier and has since joined BlueDot Impact, a nonprofit that trains people for AI safety work.

What is notable about the pair

Neither man left the field. Engels moved to an organization whose job is to measure frontier systems from the outside, and Chughtai to one that trains people to work on safety. Both chose roles that sit outside the labs while remaining pointed at them.

The timing places the departures alongside a separate development in Washington, where OpenAI said the same week that it is working with Anthropic and Google on AI safety. Pressure for outside scrutiny is now arriving from inside the labs and from policy at once, and the two paths differ on a point that matters: whether the scrutiny is funded by the companies it examines.