OpenAI's trust and safety chief steps down

Running trust and safety at a major social network is famously one of the most punishing jobs in tech. Doing it for a breakout generative AI platform seems to be even harder.

Dave Willner, OpenAI's head of trust and safety, has stepped down from his role after a year and a half. In a LinkedIn post, Willner said he is transitioning to an advisory position through the end of the year so he can spend more time with his young family, noting that the demands of the job shifted into a high-intensity phase following the explosive launch of ChatGPT.

OpenAI confirmed the move, stating that CTO Mira Murati will directly oversee the trust and safety team on an interim basis while the company searches for a permanent successor.

From Facebook's early rules to guarding DALL-E

Willner is a longtime content moderation veteran who previously led trust and safety operations at both Facebook and Airbnb.

During his early days at Facebook, he helped draft the company's first community standards, which still serve as the foundation for modern platform moderation. Back in 2009, he openly argued that hate speech was distinct from direct harm, opposing bans on Holocaust denial content on free expression grounds. His views evolved over time; by 2019, he publicly pushed back against carving out moderation exemptions for politicians.

OpenAI initially brought Willner on board 18 months ago to figure out guardrails for its image generator, DALL-E, specifically to prevent dangerous misuse like generative child sexual abuse material. But after ChatGPT landed, the scope and scale of the job ballooned almost overnight.

Bad timing on the regulatory calendar

Willner's departure comes at an awkward moment for OpenAI and the broader AI sector.

On the exact same day, OpenAI President Greg Brockman joined executives from Google, Anthropic, Inflection, Microsoft, Meta, and Amazon at the White House to sign voluntary commitments around AI safety and transparency ahead of a planned executive order. Regulators across Europe and Washington are scrutinizing generative AI like never before.

In its statement, OpenAI noted that it is now searching for a technically-skilled lead focused on the design, development, and implementation of systems to ensure safe use alongside scalable growth.

Why it matters

Dave Willner's exit highlights a brewing structural problem in AI: the old playbooks of trust and safety borrowed from Web 2.0 social networks are no longer sufficient for generative models.

In social media, trust and safety meant managing what humans posted after the fact, using policy documents, human review queues, and keyword filters. Generative AI fundamentally breaks that model because the machine itself hallucinates, invents, and generates infinite varieties of risky content in real time. Policy chops alone cannot solve this; the guardrails have to be engineered directly into the model training and system pipelines.

OpenAI explicitly calling for a technically-skilled successor is a quiet admission of this shift. As researchers from institutions like the Stanford Internet Observatory warn of looming abuse crises around generated media, the pressure on this team will only escalate. Having the department's top architect step away just as the White House demands enforceable safety pledges shows just how fragile the industry's self-regulatory armor really is.