OpenAI's Post-Training Lead for GPT-5 Defects to Anthropic as 1.5 Million Users Reportedly Flee ChatGPT
Max Schwarzer, who led post-training for the GPT-5 series, announced he's joining Anthropic — the same week reports surfaced that over 1.5 million ChatGPT users canceled subscriptions over OpenAI's defense contracts. The twin departures, one in talent and one in trust, mark a new phase of competitive pressure.
Max Schwarzer, who led post-training work across the GPT-5 model series at OpenAI, announced Tuesday that he is leaving the company to join Anthropic's reinforcement learning research team. "I've decided to leave OpenAI… joining @AnthropicAI to get back into… RL research," Schwarzer wrote, framing the move as a return to the kind of foundational research that first drew him into the field. The departure was immediately amplified across AI circles, with @cgtwts highlighting the bluntness of the announcement: a senior figure who shaped the behavior of the world's most-used AI model walking straight to its most prominent safety-focused rival.
Schwarzer's exit lands at a particularly uncomfortable moment for OpenAI. The same day, reports circulated that more than 1.5 million users have left ChatGPT in the wake of OpenAI's decision to permit U.S. defense and intelligence agencies to use its models. As @clashreport noted, many of those users have reportedly switched to Anthropic's Claude, which has publicly declined to offer unrestricted government access. @TukiFromKL put the scale in context: "Realizing ChatGPT converted like 95% of the world into AI users… and now 1.5M people are leaving it."
Get our free daily newsletter
Get this article free — plus the lead story every day — delivered to your inbox.
Want every article and the full archive? Upgrade anytime.
No spam. Unsubscribe anytime.