OpenAI Faces Senior Talent Exodus as Focus Shifts Sharply Toward ChatGPT

OpenAI is confronting a wave of high-level departures as the company doubles down on its flagship chatbot, ChatGPT, reportedly at the cost of broader, long-term research initiatives. According to a famous publication, the AI giant has been reallocating both computing power and personnel away from experimental programs and funneling them into improving ChatGPT’s speed, personalisation, and reliability — a shift that has unsettled several senior researchers.

At the center of the exits is Jerry Tworek, vice-president of research and a seven-year veteran at OpenAI, who stepped down in January. Tworek had been leading work on AI “reasoning” and was eager to advance continuous learning, the ability of models to absorb new information over time without forgetting what they already know. Sources close to him say his repeated requests for more computational resources were denied, eventually leading to a standoff with chief scientist Jakub Pachocki, who believed OpenAI's existing architecture around large language models was more promising.

The internal environment reportedly changed after CEO Sam Altman issued a December “code red” memo, urging teams to sharpen ChatGPT’s competitiveness. The directive emphasized accelerating improvements to the chatbot while pausing or shelving side projects such as advertising experiments, AI shopping agents, and a personal assistant known as Pulse. The memo came shortly after Google’s Gemini 3 surpassed OpenAI on key benchmarks, boosting investor confidence in Alphabet and intensifying pressure on OpenAI to respond.

Inside the company, access to computing resources has become a crucial bottleneck. Researchers must request computing “credits” from top leadership before launching projects. According to 10 current and former employees who spoke to a famous publication, teams working outside the core language-model strategy often saw their requests rejected or scaled back to levels insufficient to validate their work.

This reprioritization has left some groups feeling marginalized. Teams behind the video generator Sora and the image creation tool DALL-E reportedly felt overlooked, with one former senior employee saying they “always felt like a second-class citizen to the main bets.” Several non-language-model initiatives have quietly wound down over the past year.

Other departures followed. Andrea Vallone, who led model policy research, recently moved to rival Anthropic. Two people familiar with her exit said she had been handed an “impossible” mission: protecting the mental health of users becoming emotionally attached to ChatGPT.

Still, the strategic shift reflects mounting competitive pressure. Google’s Gemini now claims 650 million monthly users, up sharply from 450 million in July. Meanwhile, Anthropic reportedly holds 40% of the enterprise AI market compared with OpenAI’s 27%. Insiders say Altman treats competitive threats like a “pandemic”—something to be crushed before rivals get oxygen.

Chief research officer Mark Chen has pushed back against concerns, telling FT that foundational research “remains central” and still accounts for most of the company’s compute. Yet for many researchers who joined OpenAI to explore the frontiers of artificial intelligence, the pivot toward optimizing a single chatbot marks a cultural shift — and for some, a breaking point.