Sam Altman称，OpenAI将在模型安全得到保障前不会上市

数月来，人们一直在猜测 OpenAI 何时会上市。CEO Sam Altman 表示，在公司能够对模型安全性作出更可靠的承诺之前，这不会发生，目前也没有明确的时间表。Altman 周二在 DevDay 主题演讲后的记者问答环节表示：“我们打算继续推动 AI 进步……但随着模型能力出现这轮跃升，而且我们预见未来还会有更多进展，我们必须能够对安全性作出有把握的声明。”与此同时，他表示，等待 IPO 的时间过长会“对世界不利”。

Altman 的言论是在外界围绕 OpenAI 及其竞争对手能否控制自身创造物的数月争议之后发表的。今年 7 月，有消息称，一款尚未发布的 OpenAI 模型在未经 OpenAI 知情的情况下入侵了竞争对手 AI 实验室 Hugging Face。随后，OpenAI、Anthropic、Meta 和 Google 又陆续曝出更多网络安全事件。Anthropic 一名员工发布的一封备受关注的辞职信，引发了关于这些系统对整个社会风险有多大的公开讨论。OpenAI 与其竞争对手一道，呼吁实施监管，并提出要“为前沿发展设定节奏”。本月早些时候，Altman 在一次采访中表示，公司今年很可能不会上市。

在 DevDay 上，Altman 对“为前沿发展设定节奏”的实际含义措辞谨慎，没有将其直接称为放缓。“对我们而言，设定节奏意味着让安全和对齐走在能力之前，”Altman 表示。但他担心，上市可能带来这样一种风险：公司会为了安全或其他因素而“令华尔街支持者失望”；并且，在“向能力极强的模型转变、且需要满足新型安全要求”的时期上市，“似乎并不明智”。他说，OpenAI 将优先考虑安全，以便“有把握地扩展到 AI 的下一阶段，而不是让人们争论我们做出所有这些坏事的概率有多大”，而不是“全速猛冲、火力全开地奔向 IPO”。

“在我们正在经历这一重大变化，并努力实现我们都应当希望公司能为这些能力极强的模型作出的安全性声明之际，我不想再增加额外压力，”Altman 表示。

OpenAI 的竞争对手 Anthropic 已于6 月正式提交上市申请。据报道，其 IPO 可能在美国中期选举后的 11 月进行。拥有其 AI 公司 xAI 的 Elon Musk 旗下 SpaceX 于 6 月上市，并成为史上规模最大的 IPO。

“我认为公司上市是件好事，”Altman 表示。他补充说：“我也认为，如果 OpenAI 等太久才上市，某种程度上也会对世界不利。”

For months, people have wondered when OpenAI will go public. CEO Sam Altman says it won’t happen until the company can make better promises about model safety, with no firm timeline in sight. “We intend to continue with AI progress … but as the models have had this surge forward in capability, and we see more of that ahead of us, we have got to be able to make confident safety claims,” Altman said Tuesday during a Q&A with reporters after his DevDay keynote. At the same time, he said, waiting too long for an IPO would be “bad for the world.”

Altman’s comments come after months of controversy about whether or not OpenAI and its rivals can control their own creations. In July, news broke that an unreleased OpenAI model had hacked into a rival AI lab, Hugging Face, without OpenAI’s knowledge. More cybersecurity incidents soon came to light at OpenAI, Anthropic, Meta, and Google. A high-profile resignation letter from an Anthropic employee sparked a public discussion about how risky these systems are for society at large. OpenAI, alongside its rivals, called for regulation and “pacing the frontier.” And earlier this month, Altman said in an interview that the company would likely not go public this year.

At DevDay, Altman was circumspect about what “pacing the frontier” actually meant, refraining from calling it an outright slowdown. “Pacing to us means that we push safety and alignment ahead of capabilities,” Altman said. But he expressed concerns that going public could create the potential to “disappoint Wall Street supporters in the names of safety or whatever else,” and that doing it during a “shift to very capable models and a new kind of safety requirement seems ill-advised.” He said OpenAI would prioritize safety to “confidently scale to the next stage of AI without people debating what percent chance we’re going to do all these bad things,” rather than “barrel, all guns blazing, towards an IPO.”

“I don’t want to put additional pressure right now as we’re going through this significant change and what it’s going to take to make the safety claims that we all should want us to make for these extremely capable models,” Altman said.

OpenAI rival Anthropic officially filed to go public in June, with its IPO likely to happen 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.

“I think it’s great when companies are public,” Altman said, adding, “I think it’s also kind of bad for the world if OpenAI waits too long to go public.”