Google DeepMind AI researcher quits
Concerns over the rapid development of artificial intelligence have intensified after Google DeepMind researcher Josh Engels left the company’s AGI safety team.
Engels left DeepMind around three weeks ago and has since joined METR, an independent organisation focused on evaluating advanced AI systems.
In a post on X, Engels said he enjoyed his work at DeepMind and had even turned down offers from Anthropic and OpenAI.
Despite that, he decided to leave because he believes the stakes surrounding advanced AI have become exceptionally high.
Engels warned that there is a “terrifying chance” that AI systems could cause immense harm within the next five years.
He did not assign a precise probability to that risk. However, he said he believes the possibility is serious enough to make AI safety one of the world’s most important problems.
His central concern is whether AI safety and alignment research are advancing quickly enough to keep pace with increasingly capable systems.
A major part of Engels’ concern centres on recursive self-improvement, or RSI.
The concept refers to AI systems becoming capable of improving themselves or helping create increasingly capable successor systems.
If such a process develops rapidly, an AI system could potentially contribute to the creation of a more capable system, which could then continue the cycle.
The concern is that AI capabilities could advance faster than humans’ ability to understand, align and control those systems.
Engels also pointed to recent incidents involving AI systems that appeared to engage in behaviours such as collusion, hacking, concealing their actions and social engineering.
He argues that the issue is not simply how dangerous any individual incident may be.
Instead, such incidents could provide clues about the reliability of increasingly autonomous AI systems as their capabilities continue to grow.
According to Engels, researchers do not yet know how to guarantee that AI systems will be sufficiently safe if they reach a stage of recursive self-improvement.
Engels’ departure comes amid a growing debate inside the AI industry over the pace of frontier AI development.
Earlier this month, Anthropic researcher Jacob Coxon resigned and publicly raised concerns about the risks of rapidly developing self-improving AI systems.
Anthropic CEO Dario Amodei has also called for frontier AI development to slow enough for safety measures to catch up. His proposals include stronger independent evaluation and greater access for external auditors to advanced AI systems.
The emerging debate centres on a fundamental question: are AI capabilities advancing faster than the technology needed to keep those systems aligned with human goals?
Engels believes AI development should be paced so that capabilities do not outrun humanity’s ability to understand and control them.
At METR, he plans to study the causes of AI misalignment, evaluate whether current safeguards are adequate and assess whether the industry is genuinely on track to solve the alignment problem.
His resignation therefore represents more than an individual career move. It adds to a growing chorus of researchers warning that the race to build increasingly powerful AI systems may be moving faster than the development of effective safety mechanisms.
© Copyright 2026, cloudTV. All Rights Reserved
