❗️Andrea Vallone, who led safety research at OpenAI focused on AI responses to mental health distress and emotional over-reliance, has joined Anthropic’s alignment team.

She previously built OpenAI’s “model policy” research team and helped shape deployment strategies for GPT-4 and GPT-5, including training methods like rule-based rewards. V

allone now joins Jan Leike, OpenAI’s former safety chief who left in 2024 over concerns that OpenAI’s safety processes were being sidelined by product ambitions.

Her move comes amid growing concern over chatbot-induced mental health crises, including m