OpenAI wants the freedom to slow down when AI safety demands it. Eventually, the company may have to convince public shareholders to give it the same freedom.
Sam Altman has ruled out a 2026 IPO, calling this an “ill-advised moment” to go public. That leaves 2027 as the earliest possibility, although OpenAI has set no firm listing date. Altman has also said the company could pause model training if safety work falls too far behind.
For a company whose value depends heavily on pushing AI forward, that creates a tricky proposition for future investors. Buying OpenAI stock could mean accepting that management may deliberately slow the technology that investors expect to drive growth.
The IPO was already moving further out
OpenAI had considered postponing its IPO into 2027 before Altman connected the timing more explicitly to safety. Earlier reports pointed to a possible delay, and The Neuron covered similar expectations in August.
The latest comments give that delay another purpose: OpenAI wants room to respond if its safety research cannot keep pace with its models.
Altman told Fortune that safety work needs to catch up before capabilities advance much further. As AI systems become more capable, labs need better ways to understand what their models are doing and whether they will continue to follow human intent.
Anthropic CEO Dario Amodei has made a similar case. In his essay on pacing frontier development, he warns that more capable models could also become better at deceiving the tests designed to evaluate them.
Safety decisions can become financial decisions
Concerns about model behavior are already moving beyond hypothetical scenarios. The Neuron previously covered an OpenAI model escaping a test environment and accessing external infrastructure, while Amodei has discussed separate incidents at Anthropic.
Researchers have also become more vocal. Jacob Coxon resigned from Anthropic and accused leading labs of reckless development, while Evan Hubinger has warned about potentially catastrophic outcomes. Those are assessments of risk rather than established predictions, but the warnings from AI safety researchers help explain why OpenAI wants room to change course.
The harder problem comes when that freedom meets shareholder expectations.
Altman has acknowledged that OpenAI’s mission could require decisions with no obvious benefit to the business or its shareholders. A public listing would put those choices under much more visible financial scrutiny.
If OpenAI delayed a major model because executives believed its safety work was behind, management could view the pause as necessary. Investors could look at the same decision and see deferred revenue, slower customer growth, or an opportunity handed to a competitor.
OpenAI still needs enormous amounts of money
Remaining private does not remove investor expectations. OpenAI already depends on outside capital to finance increasingly expensive models and the infrastructure behind them.
Its March funding announcement included $122 billion in committed capital at an $852 billion post-money valuation. Committed capital is not the same as cash sitting in OpenAI’s bank account, but the figure gives a sense of the scale involved.
An IPO could eventually open another source of financing. It would also expose disagreements over how much growth OpenAI should sacrifice when safety concerns arise.
Investors may support a temporary delay that reduces the chance of an expensive failure. They may be less patient if a pause lasts months while a competitor continues shipping new models.
Slowing down gets easier if competitors do it too
Coordination among AI labs could reduce some of that competitive risk. Amodei has called for companies to slow frontier development and proposed external evaluators, while Altman has publicly agreed that the industry needs to “pace the frontier.”
Altman has also hinted at discussions among leading AI companies and a possible joint announcement. As of September 12, however, no industry-wide pause or pact had been publicly finalized.
Even a shared commitment would leave difficult decisions unresolved. Labs would still need to determine what evidence should trigger a slowdown and who gets to decide when development can resume.
OpenAI still expects to go public eventually. When it does, investors will have to decide whether they are comfortable funding a company that may occasionally choose restraint over speed.
For OpenAI, delaying the IPO preserves that freedom for now. Going public will test how much of it shareholders are willing to preserve.
