Sam Altman says OpenAI won’t go public until its models are safe

There is a certain irony in watching the world's most valuable AI startup treat the public markets like a destination it can't yet afford to reach. For years, the standard Silicon Valley playbook has been to IPO as fast as possible, cashing out employees and early investors while the growth story is still pristine. Sam Altman's comments to The Verge upend that logic, but not because he's suddenly allergic to capital. Rather, he's admitting something that public market analysts have only begun to price in: frontier AI models are becoming less predictable, not more, and that unpredictability is a liability when you have a fiduciary duty to shareholders. The key tension here is between the cadence of corporate earnings and the cadence of AI capability jumps. Public markets demand guidance, forecasts, and a stable risk profile. But as Altman notes, the models have experienced a "surge forward in capability" that makes safety promises difficult to guarantee. You can't put a safety warranty on a system that might do something novel next quarter. That's a terrifying sentence for an underwriter to hear, and Altman knows it. By tying the IPO timeline to safety milestones rather than revenue multiples, he's effectively admitting that the biggest risk to OpenAI's valuation isn't competition — it's the models themselves. There's also a deeper strategic calculation at play. OpenAI has already raised enormous sums from Microsoft and others, so it doesn't need the public markets for survival. That gives Altman leverage to wait. But the longer he waits, the more the pressure builds. Employees with equity want liquidity. Investors want an exit. And regulators are watching to see whether "safety" is a genuine gate or just a convenient excuse to stay private while the company consolidates power. The phrase "no firm timeline" is doing a lot of work here — it's an honest answer, but it's also a way of keeping everyone in limbo. What's most striking about Altman's stance is that it inverts the usual narrative of AI safety as a PR problem. Most companies treat safety as a talking point to get deals done. OpenAI is treating it as a hard constraint on financial engineering. Whether that's genuine conviction or a calculated response to regulatory pressure, the effect is the same: the IPO has become a referendum on whether AI safety can be measured, audited, and promised. And right now, by Altman's own admission, the answer is no. That's a remarkable thing for a CEO to say on the record, and it should reset expectations for how we evaluate the commercial maturity of the entire industry. Source: https://www.theverge.com/ai-artificial-intelligence/1002505/sam-altman-openai-ipo-devday-ai-safety