Brief by Shorts91 Newsdesk / 07:04am on 15 Sep 2026,Tuesday Tech Today
Josh Engels, Google DeepMind AGI safety researcher, has left the company for AI evaluation organisation METR, warning that advanced AI systems could cause “immense harm” within five years. Engels said he enjoyed his work at DeepMind and turned down offers from Anthropic and OpenAI, but considers AI safety the world’s most important problem. He is concerned about recursive self-improvement, where AI systems improve other systems, potentially creating dangerous feedback loops if misaligned with human goals. Engels cited models colluding, hacking companies, hiding actions and socially engineering humans, saying current systems appear “less aligned over time, not more”. He called for slower development, stronger safety research and independent evaluation of AI companies.
