Anthropic’s Safety Lead just quit and his goodbye letter is terrifying

I’ve seen a lot of tech departures, but this one is different. Mrinank Sharma, the head of safeguards research at Anthropic, just stepped down, and he didn’t go quietly. He posted a two-page letter on X that has the entire AI safety community on edge.

‘The world is in peril’

That isn’t hyperbole—those are his exact words. Sharma wrote that he is leaving not because of a better offer, but because he is reckoning with a ‘world in peril’ from a series of ‘interconnected crises unfolding’ right now. He wasn’t just talking about AGI going rogue; he was talking about the broader landscape of how these models are being deployed.

The world is in peril… Had I more courage, I would have left sooner.

Mrinank Sharma, Former Head of Safeguards at Anthropic

Netizens are losing it

If you check the replies on X, the reaction is a mix of genuine fear and deep philosophical debate. Some are calling it a ‘canary in the coal mine’ moment for AI safety, while others are dissecting his cryptic references to poets. One user commented, ‘When the guy whose job is to keep the AI safe says the world is in peril, maybe we should stop and listen.’

What does this mean for Claude?

Anthropic has always branded itself as the ‘Safety First’ AI company. Losing their head of safeguards under these circumstances is a massive blow to that image. It makes you wonder what’s happening behind closed doors—especially with the recent news of their Series G funding valuing them at 80B. Are we sacrificing safety for scale?

I don’t have the answers, but reading that letter made me want to go offline for a week. The fact that someone so close to the technology is this worried should make us all pay a lot more attention.