OpenAI CEO Sam Altman stated that the company currently has no specific IPO plans, and that going public is not a priority until the safety boundaries of next-generation artificial intelligence are clearly defined.

Speaking at OpenAI's Developer Day on Tuesday, Altman said an IPO may only truly come onto the agenda after the company completes "several safety case validations" and develops a clearer understanding of how to safely handle "next-level AI."

Altman indicated that he does not want to push forward with an IPO while the external environment surrounding AI development and safety remains in flux.

"I don't have a specific (IPO) timetable in mind. I think now is the time to put safety and the mission first."

According to media reports the same day, OpenAI is seeking to complete more than $30 billion in financing at a $1.4 trillion valuation, using a bridge round instead of an IPO to supplement funding.

Why was the release of GPT-6.1 Astra postponed?

When asked about OpenAI's recent decision to delay the release of GPT-6.1 Astra, Altman said it was because the model did not meet some of the company's testing standards for evaluating safety and alignment, rather than because of any specific serious incident.

Altman said the model "performed slightly worse in a few of the evaluations" OpenAI uses, and attributed the decision to "extreme caution." "There was nothing scary this time," he said.

He described postponing or restructuring a model that fails to meet OpenAI's standards as part of the normal development process.

More broadly, Altman said that as model capabilities improve, alignment, safety, monitoring, and safeguards must always remain "far ahead of model capabilities." He referred to this approach as "steering" the pace of development.

AI safety challenges go beyond engineering

Altman also pushed back against the view that AI safety issues can be fully solved through engineering alone. He pointed out that the safety systems Nvidia provides for AI are not a "complete solution." AI is more of a scientific problem than an engineering one.

"We are still in the exploration stage on how to align these models," he argued. In addition to the engineering work needed to make AI systems safer, the industry also faces challenges at the scientific level.

Regarding legal liability, Altman expects a multi-layered liability framework to emerge, depending on whether the failure originates from the model itself, the developer's method of use, or intentional misuse by the end user.

He compared this potential structure to the automotive industry — in which liability may fall to different parties depending on exactly where the problem occurs.

In addition, Altman expressed cautious optimism about discussions in Washington and international coordination, while also acknowledging that previous expectations for policy progress have not always materialized.