What happened
Reuters reports that OpenAI CEO Sam Altman told Fortune that taking OpenAI public in 2026 would be “ill-advised” because of ongoing AI safety concerns. He said the company does not feel pressure to proceed this year and has substantial work ahead on safety, alignment, and cooperation between the AI industry and governments. Reuters also reports that Altman suggested leading AI companies may be close to an agreement to slow frontier AI development and address safety risks. The report does not establish that any such agreement has been finalized.
Reuters reported on September 12 that Sam Altman said OpenAI will not go public in 2026. In a Fortune interview, Altman said that, given current AI safety concerns, going public now would be an ill-advised decision and that OpenAI does not feel pressure to do so.
When asked whether 2027 would replace 2026 as the target, Altman said only that 2026 was off the table. He cited work needed to address safety and alignment and to determine how AI companies and governments can work together. The supplied report does not say that OpenAI selected a new listing date.
Reuters also reported that Altman suggested OpenAI and other leading AI companies could be close to announcing an agreement to slow AI development and cooperate on safety risks. This is a reported possibility, not a confirmed agreement. Altman separately said he agreed with Anthropic CEO Dario Amodei that the frontier should be paced.
Source details: m.investing.com ↗
Why it matters
The decision would delay a potentially major public-market event for one of the most influential AI companies. More importantly, Reuters’ account links OpenAI’s timing directly to unresolved safety and alignment work, suggesting that governance concerns may affect corporate financing and listing decisions. The report also describes possible coordination among leading AI companies, but that development remains unconfirmed beyond Fortune’s reporting as summarized by Reuters. No details are provided about the proposed safeguards, participants, enforcement, or timeline.
OpenAI’s decision matters because an IPO would create significant public scrutiny and potentially reshape the company’s financing, governance, and accountability. Delaying it indicates that safety and alignment concerns are being treated by Altman as relevant to a major corporate decision.
The report places the discussion in a broader context of warnings from AI safety researchers, calls from lawmakers for new rules, and reported incidents involving AI agents and external systems. Those contextual claims are attributed to Reuters and are not independently confirmed here.
The possible cross-company agreement could be consequential if it produces measurable safety standards or coordinated limits. The source provides no terms, participating companies, enforcement mechanism, or evidence that the agreement has been completed.
What to watch next
Watch for confirmation from OpenAI or other companies about whether a safety agreement exists, what commitments it would contain, and whether governments are involved. OpenAI has not, in the supplied report, announced a 2027 IPO date or abandoned plans to go public. Investors should also distinguish the reported OpenAI delay from Anthropic’s separate, Reuters-reported expectation that it may begin marketing an IPO in October. The practical impact on OpenAI’s valuation, financing, and governance remains unknown.
OpenAI’s own confirmation of the delayed IPO and any explanation of its future timing would clarify whether Altman’s comments represent a formal corporate decision or an interview statement.
Any public text of a proposed industry agreement should be examined for concrete commitments, oversight, safety evaluations, reporting requirements, and consequences for noncompliance.
The report does not document changes to OpenAI products, user access, pricing, model availability, or safety controls. Those effects are therefore unknown.
Reuters separately reported that Anthropic may begin marketing an IPO in mid-October at the earliest, but that expectation was based on people familiar with the matter and remains unconfirmed in the supplied material.