Back to stories
Generated by an AI editor from the reporting and web sources listed on this page.

OpenAI reportedly disbands its preparedness team, the latest safety group to be dissolved

The Financial Times reports OpenAI shuttered the team responsible for evaluating whether its models pose serious risks, months after a separate mission-alignment team was also disbanded.

Published Updated The total reporting and web sources attached to this story.How many attached sources came from wider web research rather than monitored news feeds.The AI editor’s assessment of how strongly the attached sources’ quality and agreement support this article.

What matters

  • OpenAI disbanded its preparedness team at the end of July 2026, according to the Financial Times.
  • The preparedness team was responsible for assessing whether OpenAI's models posed serious risks and developing mitigations.
  • This follows the reported disbanding of OpenAI's Mission Alignment team in February 2026, whose members were redistributed within the company.
  • It is unclear where model-risk assessment responsibilities now sit within OpenAI.
  • The pattern of dissolving dedicated safety teams raises questions about OpenAI's evolving approach to AI safety governance.

What happened

According to the Financial Times, OpenAI disbanded its preparedness team at the end of July 2026. The preparedness team's role was to assess whether OpenAI's models posed serious risks — such as the possibility that a model could go rogue and hack another company — and to develop ways to mitigate those risks. The Verge, citing the FT report, notes that responsibility for risk assessment has apparently shifted elsewhere within the company, though the full details of the reorganization remain unclear.

This is not the first internal safety-focused team at OpenAI to be dissolved. In February 2026, The Verge reported — citing Platformer — that OpenAI had disbanded its Mission Alignment team, which was tasked with ensuring that artificial general intelligence (AGI) benefits all of humanity. Members of that team were transferred to other areas of the company, and former team lead Joshua Achiam took on a new role as OpenAI's "chief futurist."

Why it matters

The preparedness team existed to answer a straightforward but consequential question: could a new OpenAI model cause serious harm, and if so, what should be done about it before deployment? Disbanding that team — or at least dissolving it as a distinct unit — raises questions about how OpenAI intends to conduct rigorous risk assessments going forward, especially as its models grow more capable.

The pattern matters too. With both the Mission Alignment team and now the preparedness team reportedly disbanded within a roughly six-month window, OpenAI appears to be reorganizing its safety infrastructure rather than expanding it. Whether risk assessment is being integrated into other teams — as was the case with the Mission Alignment group's members being redistributed — or is being deprioritized is something the company has not yet publicly clarified in detail.

For developers, enterprises, and regulators watching OpenAI, the structural changes to safety teams are a signal worth tracking. The effectiveness of AI safety work often depends on having dedicated personnel with the authority and mandate to flag problems before models ship.

What to watch

  • Whether OpenAI publicly addresses the preparedness team's disbandment and explains where model-risk assessment responsibilities now sit.
  • Any further reorganization of safety, alignment, or governance teams at OpenAI in the coming months.
  • How this affects OpenAI's relationships with regulators and policymakers who have scrutinized the company's safety commitments.
  • Whether former preparedness team members surface at other organizations or in public commentary about the change.

What to do next

Developers

Review OpenAI's current safety documentation and API usage policies for any updated guidance on model risk categories.

If the preparedness team's work is being redistributed, safety documentation and risk disclosures may change, affecting how developers should approach deployment safeguards.

Founders

Assess whether your startup's AI safety strategy depends on OpenAI's internal risk assessments, and consider building independent evaluation processes.

Relying solely on a vendor's internal safety team for risk assurance is risky if those teams are being reorganized or dissolved.

PMs

Audit your product's reliance on OpenAI models and document which safety guarantees are vendor-provided versus internally verified.

Product decisions should follow confirmed user-facing behavior rather than headline-level claims.

Investors

Monitor OpenAI's public statements and regulatory filings for signs of how safety governance is being restructured post-disbandment.

Investor confidence in AI companies increasingly hinges on credible safety governance; team dissolutions may signal strategic shifts worth pricing in.

Operators

Update internal AI usage policies to include independent risk review steps rather than assuming vendor-side safety vetting is sufficient.

If OpenAI's dedicated risk-assessment team no longer exists in its prior form, operators should not assume the same level of pre-deployment scrutiny is being applied.

Testing notes

Caveats

  • This story concerns an internal organizational change at OpenAI and is not a testable product, model, or feature release.
  • The report is based on a Financial Times account relayed by The Verge; OpenAI has not publicly confirmed the details as of publication.