OpenAI IPO 2026 plans shelved as Altman cites AI safety work

OpenAI IPO 2026 plans are off the table after chief executive Sam Altman said the company would not list this year, citing the work required on artificial-intelligence safety and alignment. In an interview with Fortune published on September 12, Altman said a public offering would be ill-advised at this point and that OpenAI did not face pressure to proceed, Reuters reported.

Altman did not commit OpenAI to a 2027 listing. He said the company had substantial work to do on safety, alignment and cooperation between industry and governments, according to Reuters. An initial public offering is the process through which a privately held company sells shares to public-market investors for the first time.

The decision removes a 2026 listing from consideration but leaves the timetable beyond this year unresolved. The New York Times reported in June, citing three people involved in OpenAI's deliberations, that the company was leaning toward delaying an IPO until the following year. CNBC reported in August that finance chief Sara Friar had told employees a 2027 or earlier listing was possible if the business continued to improve.

Why is OpenAI delaying its IPO?

Altman's comments came alongside a widening debate over the pace at which companies should improve their most capable AI systems. Dario Amodei, chief executive of Anthropic, published an essay on September 12 arguing that developers should slow the rate of capability gains in frontier models. He said this did not mean stopping training or technical progress, but allowing time to align and safeguard systems and for outside parties to assess the work, CNBC reported.

Altman publicly endorsed the approach, writing that companies need to “pace the frontier” and that the subject had been central to recent discussions at OpenAI, Reuters and Politico reported. Elon Musk, who leads xAI, also backed Amodei's position on social media, writing that “Dario is right,” according to Politico.

The statements amount to support for a proposed approach rather than a confirmed industry agreement. Reuters reported that Altman suggested OpenAI and other leading developers might be close to announcing a pact on slowing development and addressing safety risks. No completed agreement was reported.

What would Amodei's AI pacing plan involve?

Amodei's proposal has three parts, according to CNBC. First, independent evaluators would receive employee-level access to a company to test its safety practices and report incidents. Amodei said Anthropic had already made a unilateral commitment to this step.

• Leading AI companies in democratic countries would develop shared safety standards.
• Democratic governments would seek coordination with authoritarian governments on the issue.

Politico reported that Amodei's argument is that a slower pace of capability improvement could give researchers and governments more time to put safeguards in place. It also reported divided reactions in Congress: some lawmakers described the proposal as a constructive starting point, while others said voluntary commitments and external audits would not substitute for enforceable national rules.

Anthropic's own listing plans remain separate from OpenAI's decision. Reuters reported that people familiar with the matter expected Anthropic to begin marketing an IPO no earlier than mid-October, while noting that the company had not officially disclosed a timetable.

This story draws on original reporting from CNBC.