Sam Altman and Jakub Pachocki map OpenAI’s “third phase” and IPO-day priorities

Sam Altman and OpenAI’s chief scientist Jakub Pachocki just published the company’s “third phase” plan, and it reads like a strategic pivot with a very specific thesis: capability is not the endgame. They argue the economy is beginning to reshape around AI, and the central question now is how to make advanced AI abundant, affordable, safe, useful, and easy enough for every person and organization to benefit from it. They also make a clear distinction between “frontier capability” and what comes after, writing that the bigger task is turning advanced capability into tools people can actually use to thrive.

The plan drops the same day OpenAI announced it had confidentially filed for an initial public offering, adding immediate capital-markets urgency to a vision document. OpenAI also warned that it “may be a while” before its stock hits the market, so the IPO is not a near-term liquidity event. But the filing still changes the temperature inside the company and across the AI sector: when a company is preparing for public scrutiny, “what we do next” stops being internal strategy and becomes investor-grade positioning.

In their post, Altman and Pachocki frame OpenAI’s evolution as three phases. The first was about research toward artificial general intelligence. The second was about offering products to the world and learning how people actually use them. “Now we are entering the third phase,” they wrote, and they lay out three main goals: to build an automated AI researcher, to accelerate the economy, and to give everyone on Earth a personal AGI.

That trio is doing a lot of work. “Build an automated AI researcher” signals an internal scaling move, not just a customer-facing product push. It implies OpenAI wants to systematize parts of its R&D loop so progress is less dependent on incremental human labor. “Accelerate the economy” positions the company as an infrastructure layer, a move that can appeal to regulators and policymakers who worry about AI being disruptive without being productive. And “give everyone on Earth a personal AGI” is the ambition that forces the safety and accessibility conversation to be front and center, because making something widely available at global scale is where misuse, misalignment, and unequal access become hard constraints, not theoretical concerns.

On safety and control, the post makes its lines pretty plainly. They write that powerful systems must remain safe, aligned with human intent, and subject to human control. They explicitly reject the idea that “entirely automating everything is not the future we want,” adding that it would be unfulfilling and dangerous. Their framing is that AI should help people pursue their goals, not become untethered from them. That language matters in an IPO context because public-market investors and boards tend to press for clarity around risk governance, not just research outcomes.

They also call for coordination beyond the company. The post reiterates their call for an international organization that would work on reducing risks posed by AI and could slow down the development of frontier models if needed. They argue that a good AI future cannot be one where a small number of institutions control most of the capability and most of the upside, and they describe a future where many people, companies, communities, and countries can build, benefit, and hold power. That is both a policy argument and a business argument. If the market believes AI is a winner-takes-most industry, then a public OpenAI needs a narrative that explains why the upside is not only captured by a handful of labs.

The IPO timing also invites immediate comparison to other frontier labs. The Business Insider piece notes the blog post was reminiscent of a separate post from Anthropic, where researchers said AI is advancing so quickly that leading frontier labs may need to slow down. In that Anthropic post, the researchers argued it would be good for the world to have the option to slow or temporarily pause frontier AI development so societal structures and alignment research can keep up with the technology. Anthropic quoted, “We believe it would be good for the world to have the option to slow or temporarily pause frontier AI development to enable societal structures and alignment research to keep up with the advance of the technology.”

So what should executives watching OpenAI take from this? First, Altman and Pachocki are trying to define the next competitive frontier as distribution and usability, not just model performance. Second, they are tying that push to safety, alignment, and human control, which signals that their third phase will include governance as a product requirement, not an afterthought. Third, the IPO filing elevates the strategic stakes. Even if the stock may not hit the market for a while, the filing process and future public scrutiny will pressure leadership teams to translate vision into measurable execution. If you’re a board member or operator at a peer company, this is a reminder that “AI abundance” is not just a slogan. It is a plan that forces every major decision about scaling, accessibility, and safety to become the same roadmap investors will follow.