OpenAI is reportedly in talks to raise $30 billion at a $1.4 trillion valuation, which would be the company’s last private round before a now-delayed 2027 IPO. But CEO Sam Altman threw cold water on any near-term public debut Tuesday, saying the company won’t list until it can make better promises about model safety.

“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 during a Q&A with reporters after the company’s DevDay keynote.

There’s no timeline. That’s the actual news here.

The $1.4 trillion number is absurd in context. It would make OpenAI worth more than Amazon, more than Alphabet, and roughly equal to Apple’s current market cap. For a company that still loses money and has yet to prove it can turn its technology lead into durable profits, it’s a bet that the entire economy is about to be rebuilt around large language models.

The $30 billion raise would take OpenAI’s total funding to something north of $170 billion across all rounds. For comparison, Meta’s entire market cap in 2012, the year of its IPO, was around $104 billion. OpenAI is raising more in a single round than most companies are worth.

Altman’s safety comments are notable because they represent a real constraint, not just PR. The company has been under intense pressure from researchers, regulators, and its own employees about the risks of increasingly capable models. A dozen AI researchers, including former OpenAI and Google DeepMind employees, released interviews this week warning about existential risk. Geoffrey Irving, who worked at both companies, put the chance of human extinction at “about a coin flip.”

OpenAI doesn’t appear to be slowing down while it figures out safety. The company announced Dots at DevDay, its answer to Meta’s Muse: always-on AI agents that run in the background, connect to your apps, and learn your preferences over time. It also launched what amounts to an office suite inside ChatGPT, putting it into direct competition with Microsoft, one of its largest investors and partners.

The IPO delay also gives OpenAI more room to navigate its messy corporate structure. The company is still technically a capped-profit subsidiary of a nonprofit, a setup that made sense when it was a research lab but looks increasingly strained as it chases a valuation larger than most countries’ GDP.

Investors are clearly willing to wait. Raising $30 billion at that valuation means someone believes OpenAI is worth the risk, the timeline uncertainty, and the governance complexity. It also means the company can keep operating as a private entity for longer, without quarterly earnings calls or the scrutiny that comes with being a public company.

But Altman’s comments suggest the IPO isn’t just delayed for logistical reasons. It’s delayed because the company doesn’t yet know how to promise regulators, investors, and the public that its models won’t cause catastrophic harm. That’s not a problem you solve with better PR. It’s a technical and political challenge that could take years.

The funding round, if it closes, will be the largest in startup history. The IPO timeline is now anyone’s guess.

The other OpenAI news that matters

OpenAI skipped Nvidia’s Open Agent Safety Platform, a new industry effort to prevent rogue AI agents. The company isn’t a public supporter, but TechCrunch learned it’s working with Nvidia privately. That’s worth noting because OpenAI’s absence from public safety efforts has become a pattern. The company talks about safety constantly but often operates alone.

Protesters showed up at DevDay. More than a dozen organizations, including Bay Resistance, Tech Workers Coalition, and Service Employees International Union, rallied outside the venue at San Francisco’s Fort Mason. The chant: “Sam Altman, get off it, put people over profit.” The protest focused on OpenAI’s contracts with ICE and the environmental cost of its data centers.

The security money keeps flowing

Reco, an AI agent security startup, raised $55 million, bringing its total funding to $140 million. The round came just seven months after a $30 million raise in February, which tells you how much demand there is for tools that secure AI agents before they become a liability.

The market is crowded. Every major security vendor is building agent-specific products, and a dozen startups are chasing the same problem. Reco’s bet is that companies will need dedicated tools to manage the permissions, data access, and behavior of AI agents as they proliferate across enterprises.

It’s a reasonable bet. Agents like OpenAI’s Dots and Meta’s Muse are designed to connect to your apps, access your data, and act on your behalf. That’s useful, but it’s also a massive attack surface. Reco’s timing is good. The technology is moving faster than anyone’s ability to secure it.