When OpenAI discovers a social conscience

Sam Altman just published a thirteen-page manifesto to reshape society in the face of AI. Right before the IPO. Two years after disbanding his safety team.

---

Picture the CEO of Marlboro publishing an urgent report on the harms of tobacco a few weeks before taking the company public. Absurd? Something close to that just happened in the AI world.

In early April 2026, OpenAI published a thirteen-page document titled “Industrial Policy for the Intelligence Age: Ideas to Keep People First.” The title sounds gentle. The substance is genuinely radical: nothing less than rethinking the economic and social foundations of our societies to absorb the shock of superintelligence. Four-day work weeks, taxes on robots, a sovereign fund distributed to every citizen, automatic safety nets triggered by economic thresholds. Altman compares it to Roosevelt’s New Deal. Nothing modest about it.

The paper is signed OpenAI, but Altman’s fingerprints are everywhere. That is precisely where it gets interesting.

What the paper says — and what it leaves out

It is only fair to acknowledge that the proposals are not hollow. Some are genuinely ambitious.

The headline idea: a public sovereign wealth fund, partly funded by AI companies themselves, that would give every citizen a share of the economic growth generated by the technology. A universal technology dividend. In parallel, the document anticipates that AI will erode traditional tax bases — less payroll, fewer contributions — and proposes shifting taxation toward capital gains and corporate profits. There are also economic safety nets designed to activate automatically when job-displacement indicators cross certain thresholds, without waiting for a vote in Congress. And in what the document itself calls a “chilling” passage, OpenAI acknowledges scenarios where dangerous AI systems could not “easily be recalled” because they would be autonomous and capable of self-replication.

That is a remarkable admission from the people building the thing.

Here is the rub. This paper does not land in a vacuum. It lands at the exact moment OpenAI is preparing to go public, under pressure from investors and competition. Positioning itself as the “responsible” actor in AI at that precise instant is no accident.

The gap between words and deeds

There is one thing this document cannot erase: the track record.

OpenAI actively pushed to weaken oversight mechanisms in the European AI Act. In California, when bill SB 1047 proposed risk-management strategies very close to what Altman himself had asked Congress for, OpenAI opposed it. More recently, the company has been accused of using intimidation tactics to sink the California Transparency in Frontier AI Act.

The most telling story remains that of the “superalignment” team, disbanded in 2024. OpenAI had promised it 20 percent of its compute for safety work. It ultimately received only 1 to 2 percent, on aging hardware. Co-lead Jan Leike left publicly, citing Altman’s abandonment of the safety mission. The team was dissolved. Its successor lasted sixteen months.

Two years later, in this new document, OpenAI proposes that this safety work should now be funded by governments. The formula is simple: externalize the costs, keep the profits.

Who names the problem writes the rules

There is a well-known dynamic in politics and lobbying: being the first to name a threat is a way to claim the right to define its contours — and therefore what counts as an adequate response. If you decide what is dangerous, you also decide what counts as enough.

That is the game OpenAI is playing here at global policy scale. By publishing this manifesto, Altman is essentially saying: we see what is coming, we are here to help you through it, trust us to steer the process.

The problem is that the ideas on offer are not new. Sharing the benefits of AI, mitigating risks, democratizing access — that has been the chorus of every serious conversation since ChatGPT launched. The real challenge is not naming those goals; it is building concrete mechanisms to achieve them. OpenAI proposes to “start the conversation” and offers up to $100,000 in cash and $1 million in API credits to fund research aligned with its proposals. That is generous. It is also barely binding.

Some experts welcome the document as a useful contribution — others see well-dressed regulatory nihilism: wave reforms so ambitious they cannot pass, while blocking more modest reforms that could apply tomorrow morning.

What we can still take from it

For anyone tracking AI’s impact on the labor market, the document is worth reading — not as an action plan, but as an industrial admission.

When the world’s leading AI company states in black and white that superintelligence will drive a transformation as deep as the industrial revolution, that current safety nets are inadequate, and that jobs will be displaced faster than institutions can react — those statements carry weight on their own, whatever the author’s motives.

The real question is no longer whether Altman is sincere. It is whether governments will settle for being invited to the conversation OpenAI wants to host — or whether they will write the rules themselves, without asking permission.

For now, OpenAI has achieved one thing: placing itself at the center of the debate. That is already something.

---

Sources: OpenAI, “Industrial Policy for the Intelligence Age” (April 2026); Tech Policy Press; Fortune; The Hill; Gizmodo.