OpenAI reportedly disbanded its preparedness team as part of a 'streamlining' process

OpenAI has reportedly eliminated its preparedness team, which was responsible for evaluating catastrophic risks posed by its AI models, as part of a broader organizational restructuring effort. The move raises concerns about whether safety oversight is being deprioritized as the company scales up operations.

OpenAI has reportedly shut down an internal group dedicated to identifying and assessing catastrophic risks associated with its artificial intelligence models, folding the team as part of what the company describes as a streamlining initiative. The preparedness team was specifically tasked with anticipating worst-case scenarios — including the potential for AI systems to be misused for large-scale harm — before those risks could materialize. The disbanding comes at a particularly sensitive time, as AI capabilities are advancing rapidly and public scrutiny over safety practices at leading AI labs is intensifying. Critics worry the decision signals a shift in priorities away from proactive risk management. OpenAI has faced ongoing questions about its internal safety culture, including high-profile departures of safety-focused personnel in recent months, making this latest move likely to fuel further debate about how seriously the company treats existential and catastrophic risk.

OpenAI has reportedly eliminated its preparedness team — an internal unit charged with evaluating whether its AI models could contribute to catastrophic outcomes — as part of a company-wide restructuring described internally as a streamlining process. The group's mandate was broadly focused on anticipating and measuring extreme risks, such as AI being leveraged to develop biological weapons or carry out other large-scale societal harms, before such scenarios could emerge in the real world.

The timing of this decision is striking. OpenAI is simultaneously pushing toward more advanced and capable AI systems, commercializing aggressively, and transitioning to a for-profit corporate structure. Each of those moves independently invites questions about whether financial imperatives are beginning to outweigh safety commitments. Together, they paint a picture of an organization undergoing significant cultural and structural change.

This is not the first sign of tension within OpenAI around safety priorities. Over the past year, several prominent researchers with explicit safety mandates have departed the company, some citing concerns about the direction leadership was taking. The dissolution of the preparedness team adds another data point to that pattern.

Why it matters: Preparedness-style functions exist precisely because the most dangerous AI risks are, by definition, ones that haven't happened yet. Disbanding the team responsible for modeling those risks doesn't mean the risks disappear — it means fewer resources are dedicated to spotting them early. For an organization that has publicly committed to ensuring AI benefits humanity, removing a layer of catastrophic risk oversight sends a contradictory signal.

Broader context also applies here. Governments and regulators around the world are in the early stages of crafting AI safety requirements, and they often look to leading labs like OpenAI as benchmarks for responsible practice. If OpenAI scales back internal safety infrastructure while simultaneously lobbying against heavy-handed regulation, it risks undermining the credibility of the entire industry's self-governance argument. Observers and policymakers would be right to ask: if even the most high-profile AI safety lab is trimming its risk assessment teams, who exactly is watching for catastrophic outcomes?