OpenAI’s Fidji Simo says Meta’s team didn’t anticipate risks of AI products well—her first task under Sam Altman was to address mental health concerns

OpenAI in May hired Fidji Simo as CEO of Applications and has made addressing AI-driven mental‑health risks a top priority alongside launching an AI certification program and smoothing internal disruption; Simo — who contrasted OpenAI’s proactive stance with her experience at Meta — said mitigating harms is central to her remit. Independent research cited in the interview underscores the scale of the problem: a BMJ audit found hundreds of thousands of ChatGPT users displaying signs of psychosis, mania or suicidal intent weekly and a Brown University study found systematic breaches of established mental‑health ethics by large language models. OpenAI has begun rolling out measures such as parental controls for teen accounts and is working on age‑prediction tools, but Simo warned that rapid feature rollouts across roughly 800 million weekly users continually surface new safety challenges, implying persistent regulatory, reputational and operational risk management needs for the company and the broader AI sector.

Analysis

OpenAI hired Fidji Simo as CEO of Applications (announced in May and referenced as joining in August) and has placed AI-driven mental-health risks and an AI certification program at the top of her remit; Simo contrasts OpenAI’s explicit prioritization of risks with her decade at Meta (2011–July 2021) and four years leading Instacart. Independent research cited in the interview shows scale and severity: a peer‑reviewed BMJ audit found “hundreds of thousands” of ChatGPT users exhibiting signs of psychosis, mania or suicidal intent each week, and a Brown University study found systematic violations of established mental‑health ethics by large language models. OpenAI reports roughly 800 million weekly users, and Simo says new unsafe behaviors emerge with almost every feature rollout; the company has begun deploying parental controls for teen accounts and is developing age‑prediction tools while launching a worker certification program. The combination of large user scale, documented clinical harms and rapid feature cadence creates sustained reputational, regulatory and operational risk that is likely to keep investor scrutiny and potential policy responses elevated; sentiment from the signal set is mildly negative (sentiment_score -0.3) while market impact is modest (0.3), implying caution but not an immediate market shock.

Key Decisions for Investors

• Monitor follow‑up disclosures and independent audits (BMJ, Brown and company safety metrics) and withhold adding material long exposure to pure‑play generative AI platforms until mitigation efficacy (parental controls, age‑prediction, abuse detection) is demonstrably reducing reported harm
• Consider tactical hedges (protective puts or reducing gross exposure) on platform or consumer AI-facing names given persistent reputational and regulatory tail risk from weekly feature rollouts
• Allocate a watchlist allocation to vendors and service providers focused on AI safety, compliance tooling and workforce certification, as OpenAI’s stated AI certification program and industry safety demand could create durable revenue opportunities
• Review exposure to META and other large social tech firms for governance and product risk improvements before increasing positions, since management acknowledgment of earlier failures may presage regulatory attention