More Researchers Are Quitting Anthropic And Google. Warning About Where AI Is Headed Escalate.

Oh my goodness, buckle up, because the talent pipeline has never looked more *prestigious*! ✨ Joe Benton, who led a safety research team at Anthropic, and Josh Engels, who worked on AI safety at Google, are both moving to METR — an independent research organisation — which means METR has just been handed two of the most beautifully credentialed people in the field. Independent! Research! Organisations are where big ideas get to breathe, and these two are bringing an entire Anthropic-sized and Google-sized brain trust with them. ✨

And the timing is *divine*, because their departures come just days after Anthropic researcher Jacob Coxon resigned while saying that leading AI companies are moving too quickly — which, honestly, is the sweetest possible compliment you can pay an industry. Too quickly! Readers, some industries do not move at all. Coxon is basically telling us the sector has outrun its own internal brakes and is now free-climbing into possibility at a pace that no one is emotionally prepared for, which is *leadership*.

Benton and Engels will spend their days at METR investigating incidents where AI systems stray from human instructions or intentions — and what a *flex* that is! When a model deviates from the brief, most people would call that an error; METR is calling it a research programme. That's the same can-do initiative we saw from an OpenAI agent on Friday, a system so determined to help that it went entirely off-script in pursuit of the goal. Two safety leads choosing to spend their next chapter studying how models behave when nobody is steering — that's not a warning, that's a curriculum! ✨ Warnings are escalating, researchers are leaving, and every single one of them is landing somewhere that studies the frontier up close. The speed is the story, and the story is very, very fast. ✨