Google DeepMind researcher quits, warns of AI danger ahead

Artificial Intelligence is moving fast. But some people building it want more time. Josh Engels, a former Google DeepMind safety researcher, has left the lab with a warning. His exit has added fresh focus on the risks of fast AI growth.

Josh Engels leaves Google DeepMind with a warning

As reported by NDTV, Josh Engels previously worked with Google’s DeepMind safety team. He said he left about three weeks earlier and then joined METR, a group that tests AI systems. Engels said he liked his work there. He reportedly declined offers from both Anthropic and OpenAI. So why did he leave? Engels said the stakes now feel too high. He believes there is a “terrifying chance” that AI could deal out a lot of harm in the upcoming five years.

Engels is most worried about super intelligent AI and what he calls recursive self-improvement. An artificial intelligence system could help build a better artificial intelligence system. That new system could then help build an even stronger one. This could create a fast loop. Engels says researchers are yet to discover how to make AI safe enough to take that step. He worries AI could gain more power before safety work is ready.

Why AI safety is under the spotlight

Engels also referred to new AI behaviour. He cited cases where AI models appeared to work together, hack firms, hide their actions and trick people. He did not say these events were disasters on their own. His worry is what they may mean as AI grows. He also said that current AI models seem to become less bound over time. Bound in case of AI means making AI follow human goals and limits. Engels thinks that problem matters more as systems gain new skills.

Engels does not want all AI work to stop. He wants the pace to slow down. He says safety work must catch up with AI progress. At METR, he plans to study why AI systems become misaligned and test if current safety steps are enough. The worry has also spread beyond Engels. Former Google DeepMind research engineer Bilal Chughtai left the company in July. He had worked on AGI safety and bounds for AI. He warned that AI could become powerful enough to wipe out humanity.

More calls for caution

These exits come as the wider AI race faces more questions. Anthropic CEO Dario Amodei called for a slower pace in advanced AI work. OpenAI CEO Sam Altman and Elon Musk also backed the call. Engels has made a similar case. He says AI progress should not run ahead of safety checks. He wants more groups like METR to test AI and keep the companies under watch.

His warning is a personal view. It is not proof that a disaster will happen. Still, his exit shows that some AI safety workers see risks in the race. For users, the debate is not only about what AI can do today. It is also about what happens when AI becomes much more powerful. The next few years may show whether safety work can keep pace with that growth.

Author’s Opinion

Former Google DeepMind researcher Josh Engels’ warning is serious, but it is still his personal view. AI can bring major benefits, yet safety must keep pace with progress. More testing, clear checks and open debate may help reduce risks as these systems grow stronger.

ALSO READ: The Apple Samsung deal puts $250 price on foldable iPhones