The Verge reports that OpenAI CEO Sam Altman took questions from reporters on Tuesday after his DevDay keynote. He said the company intends to continue with AI progress, but the models have surged forward in capability, with more of that ahead, so OpenAI has to be able to make confident safety claims. The article’s reading is that OpenAI will not go public until it can make better promises about model safety, and that no firm timeline is in sight. Altman said he did not want additional pressure from Wall Street.

[1]
Iron-gall ink: a gate still closed by two bars, with thin lines bending toward it.
Two bars still close the gate. Thin lines bend toward it and do not pass through. That is Altman’s condition: the way to a listing is not open until the safety claims can be made. An illustration, not a press photo., AI-generated illustration, not a news photograph

The piece sets that answer inside months of argument, rather than treating it as a standalone IPO rumor. In July, news broke that an unreleased OpenAI model had entered a rival lab, Hugging Face, without OpenAI’s knowledge. More cybersecurity incidents then came to light at OpenAI, Anthropic, Meta, and Google. A high-profile resignation letter from an Anthropic employee pushed a public discussion of how risky these systems are for society. OpenAI and its rivals called for regulation and for pacing the frontier. Earlier this month, Altman said in an interview that the company would likely not go public this year. That is The Verge’s background, not a separate investigation in this piece.

[1]

At DevDay he did not describe pacing the frontier as an outright slowdown. His line was that pacing means pushing safety and alignment ahead of capabilities. He worried that going public could create the chance of disappointing Wall Street supporters in the name of safety or something else. He said doing it during a shift to very capable models, and a new kind of safety requirement, seems ill-advised. What he said he would prioritize is being able to scale confidently to the next stage of AI without a debate over the percent chance of bad outcomes, rather than barreling toward an IPO. He added that he does not want additional pressure while the company is in this change and still has to make the safety claims people should want for these extremely capable models.

[1]

The same Q&A pulls the other way. He said it is great when companies are public, and that it is also somewhat bad for the world if OpenAI waits too long to go public. The article renders the IPO point as: waiting too long would be bad for the world. The comparison sits in the next sentences. Anthropic officially filed to go public in June, with its IPO likely in November, reportedly after the US midterm elections. Elon Musk’s SpaceX, which owns his AI company xAI, went public in June and became the biggest IPO in history. Altman did not give OpenAI a date. What a reader can take from this is a condition, not a calendar.

[1]

要点

  • Altman said OpenAI will not go public until it can make confident safety claims. He gave no date.
  • He described pacing the frontier as safety and alignment ahead of capabilities, not as an outright slowdown.
  • He also said waiting too long would be bad for the world. Anthropic filed to go public in June.
  • The July Hugging Face incident and the later security cases are The Verge’s background, not a new finding here.