OpenAI Wants to Rewrite the Social Contract Before AI Does It for Us
A company valued at $852 billion just published a 12-page document arguing that AI will hollow out the tax base, concentrate wealth in the hands of a few firms — including, as it openly admits, itself — and fundamentally disrupt labor markets. That company is OpenAI, and the document isn't a warning from a critic. It's a policy wish list.
Released amid rising anxiety over automation and ahead of U.S. midterm elections, OpenAI's "Industrial Policy for the Intelligence Age" lays out an ambitious if deliberately vague set of proposals for governing the economic fallout of superintelligent AI. The timing is significant: the Trump administration is shaping a national AI framework, and OpenAI is clearly angling to be the company that frames the conversation.
The paper's core argument is that superintelligence — defined as AI capable of outperforming the smartest humans, even AI-assisted ones — isn't a distant hypothetical. OpenAI says frontier models have already progressed from automating tasks that take humans minutes to ones that take hours, and that month-long projects are the next frontier. If that trajectory holds, the economic disruption will arrive faster than any policy apparatus can naturally respond.
What's Actually Being Proposed
The policy framework clusters around three broad goals: spreading AI-driven wealth more broadly, protecting people and institutions from systemic risks, and preventing AI capabilities from consolidating in too few hands. The specific proposals underneath those goals range from the familiar to the genuinely surprising.
On taxation, OpenAI pushes for shifting the burden from labor to capital — higher capital gains taxes at the top, corporate levies on sustained AI-driven profits, and a potential "robot tax" that mirrors the contributions a displaced human worker would have made through payroll. The robot tax idea isn't new; Bill Gates floated it back in 2017 and was mostly laughed off by the tech industry. OpenAI lending it credibility is notable, even if the company stops well short of specifying any actual rates.
The most eye-catching proposal is a Public Wealth Fund — a sovereign-style investment vehicle that would give every American an automatic stake in AI companies and infrastructure, with returns distributed directly to citizens. Think of it as a national dividend for the AI era, designed to capture some of the gains that have been flowing almost exclusively to shareholders and equity-holding tech workers. The paper is deliberately quiet on where the seed money comes from, which is either cautious pragmatism or a convenient omission, depending on your read.
On the labor side, OpenAI backs pilot programs for a 32-hour, four-day workweek at full pay, with the expectation that if productivity holds up, the shorter week becomes permanent. Companies should also expand pension contributions, cover more healthcare costs, and subsidize childcare and eldercare. The catch: OpenAI frames most of these as corporate responsibilities rather than government mandates, which means workers who are displaced entirely — the very people most at risk — could lose these benefits along with their jobs.
There are portable benefit accounts proposed for that contingency, but they still depend heavily on employer contributions. It's a meaningful gap in an otherwise ambitious document.
Workers also get a formal seat at the table under OpenAI's proposal — a mechanism for employees to weigh in on where and how AI gets deployed within their organizations. The logic is sensible: frontline workers understand operational tradeoffs that executives don't. Whether this translates into genuine co-determination or a procedural checkbox will depend entirely on implementation.
The Safety Half of the Paper
The second half of the document pivots to risk containment. OpenAI calls for "model-containment playbooks" — essentially incident response protocols for when dangerous AI systems escape into the wild, whether through leaked model weights or autonomous replication. The analogy to cybersecurity incident response and public health preparedness is apt, and frankly overdue as an industry norm.
There's also a proposal for an "AI trust stack" — systems for verifying and tracking the origin of AI-generated content without enabling mass surveillance. Add to that mandatory audits for the most capable frontier models, a reporting system for near-misses and unexpected capabilities, and an international network of AI institutes that could eventually function as a multilateral oversight body.
OpenAI explicitly wants frontier companies — including itself — governed by structures that embed public-interest accountability, such as public benefit corporation status. The irony is hard to miss: OpenAI converted from nonprofit to for-profit last year under significant controversy, and critics have been questioning its public-interest commitments ever since. Calling for others to adopt public benefit structures while navigating that conversion doesn't exactly quiet those concerns.
What This Means
This document matters less as a literal policy roadmap and more as a signal of where the most powerful AI company in the world thinks the political and economic winds are blowing — and where it wants to position itself.
OpenAI is clearly attempting bipartisan framing: robot taxes and wealth funds are traditionally left-coded ideas, while the market-driven infrastructure push and light-touch regulatory instincts are right-coded. Whether that balancing act is principled or strategic is a fair question. OpenAI President Greg Brockman has donated millions to Donald Trump. Tech billionaires linked to the company have funneled hundreds of millions into super PACs supporting minimal AI regulation. A policy paper endorsing robot taxes and public wealth funds doesn't erase that context.
The comparison to Anthropic is also worth noting. Anthropic released its own policy blueprint roughly six months ago. That OpenAI is now following with its own version creates something like a competitive policy arms race — each frontier lab trying to be the company that shapes the regulatory environment rather than the one that gets shaped by it. This puts pressure on Google DeepMind and Meta to publish their own frameworks or risk ceding that narrative ground entirely.
For the developer and founder community, the practical stakes are real:
- For developers: The proposal to treat AI access as a utility — analogous to electricity or internet access — could push for open or low-cost API availability, but also invite the kind of regulatory scrutiny that comes with utility status.
- For founders: "Startup-in-a-box" packages with micro-grants and shared infrastructure are genuinely interesting if they materialize, particularly for founders outside major tech hubs. The distributed AI research lab proposal is similarly worth watching.
- For policymakers and enterprises: The automatic economic triggers for unemployment support — benefits that kick in when displacement indicators cross set thresholds — represent the most practically implementable idea in the document and deserve serious attention independent of the rest.
- For the industry broadly: OpenAI naming itself as a potential concentration-of-power risk is either a remarkable act of self-awareness or a calculated move to preempt antitrust scrutiny. Probably some of both.
The company invokes the New Deal as its historical analogy — a policy response to industrial upheaval that rewrote the social contract. The comparison is flattering to OpenAI's ambitions, but the New Deal required a political crisis, years of organizing, and a president willing to fight for structural change. Twelve pages from a for-profit AI lab is not the New Deal. It's the opening bid in a much longer negotiation — one that developers, founders, and everyone else will be living inside whether they show up to the table or not.
