Altman Offered the Government 5% of AI

How OpenAI's sovereign wealth fund proposal turns regulators into shareholders, and why the government's three-week crackdown on frontier models just got a $42.6 billion response

Author

Justin Wright
July 06, 2026 • Est. Reading Time: 6 minutes

Happy Monday!

Sam Altman discussed the proposal directly with President Trump, Commerce Secretary Lutnick, and Treasury Secretary Bessent. (source: CNBC)

On July 2, the Financial Times reported that Sam Altman proposed giving the US government a 5% stake in OpenAI, worth roughly $42.6 billion at the company's $852 billion valuation. The stake would flow into a sovereign wealth fund modeled on Alaska's Permanent Fund, which has distributed oil revenue dividends to Alaskan residents since 1982.

Altman is not proposing this for OpenAI alone. He wants every leading US AI company, including Anthropic, Google, and Meta, to cede similar stakes through the same fund. The Alaska Permanent Fund for the intelligence age.

The timing is not accidental. In the three weeks before the proposal, the US government pulled Anthropic's Fable 5 off the market, delayed OpenAI's GPT-5.6 launch, and established a de facto approval process for frontier model releases that did not exist a month ago. Altman's offer is a response to a government that just demonstrated it can ground your best model with a single letter. The question is whether this is a bribe, a peace offering, or something genuinely new.

Sam Altman proposed giving the US government a 5% stake in OpenAI through a sovereign wealth fund, and wants every major AI lab to do the same. The proposal came days after the government delayed GPT-5.6's launch by requiring individual vetting of every partner organization. In three weeks, the government pulled Fable 5, gated GPT-5.6, and prompted the most significant restructuring proposal in AI industry history. Anthropic countered with a "digital dividend" funded by future AI-sector taxes rather than equity. The two proposals reveal fundamentally different visions of the government-AI relationship.

TL;DR

The Three-Week Crackdown

The context for Altman's proposal is a government that moved faster than anyone expected.

June 12: Commerce Secretary Howard Lutnick sends Anthropic a letter designating Fable 5 and Mythos 5 as subject to export controls. Both models go offline. First government-forced takedown of a publicly deployed frontier model.

June 25: Lutnick calls Altman to warn him against releasing GPT-5.6 without prior approval from government agencies. First time the US government has preemptively intervened in a model launch before release.

June 26: OpenAI releases GPT-5.6 (Sol, Terra, and Luna) to approximately 20 organizations, each individually vetted by the government. No public access, no waitlist, and no ChatGPT support.

July 1: Fable 5 returns after 18 days offline. Anthropic retrained its cybersecurity classifier, had the fix validated by NIST CAISI, and co-developed a jailbreak severity framework with Amazon, Microsoft, and Google.

July 2: Altman proposes the 5% stake.

In less than a month, the government asserted a de facto approval process over frontier model releases at both of the two most valuable AI companies. No legislation authorized it and no regulatory framework governs it. Commerce Secretary Lutnick used export controls and phone calls.

The Three-Week Sequence

Date

Event

Significance

June 12

Fable 5 pulled (export controls)

First government-forced model takedown

June 25

GPT-5.6 launch blocked (Lutnick call)

First preemptive government intervention

June 26

GPT-5.6 released to ~20 vetted partners

Government-gated model release

July 1

Fable 5 restored (NIST validation)

Jailbreak severity framework created

July 2

OpenAI proposes 5% government stake

Labs begin negotiating terms

The Alaska Fund for AI

The proposal borrows from a proven model. Alaska's Permanent Fund was established in 1976 to invest surplus oil revenue. It is valued at roughly $91 billion and pays an annual dividend to every Alaskan resident. Altman's pitch: AI is the new oil, and the public should own a share of the extraction.

The mechanics, as reported, would work as follows: OpenAI donates (not sells) 5% of its equity into a public wealth fund. Other AI companies contribute matching stakes. The fund distributes returns, whether dividends or eventual IPO proceeds, to Americans.

Altman discussed the idea directly with President Trump, Commerce Secretary Lutnick, and Treasury Secretary Bessent. The talks remain preliminary, and any formal implementation would likely require congressional approval.

The financial scale is significant. Five percent of OpenAI at its $852 billion valuation is $42.6 billion. Five percent of Anthropic at $965 billion is $48.3 billion. Five percent of Alphabet's AI division, Meta's AI business, and any other qualifying lab would push the fund well into the hundreds of billions. This is not a token gesture. It is the largest proposed transfer of private technology wealth to the public in American history.

Anthropic's Counter

Anthropic's response reveals a fundamentally different calculation. Rather than giving the government equity, Anthropic proposed a "digital dividend" funded by future AI-sector taxes. The distinction matters.

Equity gives the government a financial interest in the company's growth. A government that owns 5% of every AI company is a government incentivized to let those companies grow, and that is exactly the point. Altman is offering to align the government's financial interests with the industry's commercial interests; regulators become shareholders.

Anthropic's tax-based approach does the opposite. It shifts the cost to after profitability. No dilution, no government ownership, no structural incentive for the government to favor the industry's growth over the public's safety. Anthropic has said it expects to be profitable by 2029, so the digital dividend would begin then.

The two proposals map onto the companies' broader strategies. OpenAI, which is losing market share (ChatGPT fell below 50% for the first time) and revenue ground (Anthropic's $47 billion run rate nearly doubles OpenAI's $24-25 billion), needs a regulatory shield. Anthropic, which is winning on revenue and enterprise adoption, can afford to wait.

Neither company denied the fundamental premise: the government is now a counterparty in frontier AI, and some form of value transfer is coming. They disagree on the mechanism, not the direction. That consensus is itself the story.

What This Means for Practitioners

For founders and startup leaders, the 5% proposal signals that building a frontier AI company now comes with a potential equity tax that did not exist six months ago. If the fund becomes reality, any company reaching frontier capability will face pressure to contribute. Factor that into your cap table planning.

For enterprise buyers, the government-gated release of GPT-5.6 creates a new procurement variable. The most capable models may not be available to you on launch day. They may not be available to you at all without government clearance. The roughly 20 organizations approved for GPT-5.6 and the approximately 100 cleared for Claude Mythos represent an emerging tier of AI access that tracks government relationships, not just willingness to pay.

For anyone watching the industry, the speed of the shift is the story. Six weeks ago, AI companies launched models whenever they wanted. Today, two of the three most valuable AI companies have had model launches blocked or gated by the government, and both are now negotiating the terms of government participation in their businesses. This happened without a single piece of AI legislation being signed into law. Colorado's attempt at comprehensive AI regulation was blocked by an xAI lawsuit with Department of Justice support. The federal government is not legislating AI. It is governing it through executive action, and the labs are negotiating directly with cabinet secretaries.

The Bottom Line

In three weeks, the US government demonstrated that it can pull a frontier model off the market, delay a model launch, and establish a de facto approval process for who gets access. It did this with export controls and phone calls, not legislation.

Altman's 5% proposal is the first serious attempt to formalize that relationship and turn regulators into shareholders. Whether that aligns incentives or captures the regulator, the offer is now on the table. The era of AI companies launching whatever they want, whenever they want, to whoever they want, ended sometime in the past three weeks.

In motion,
Justin Wright

If the government owns 5% of every frontier AI company, does it become a better regulator because it understands the technology's value, or a worse one because it profits from the technology's growth, and is there a version of this that achieves both?

Food for Thought

Quick Hits

  • Fable 5 returned globally on July 1 after 18 days offline. Anthropic, Amazon, Microsoft, and Google co-developed a jailbreak severity framework that scores attacks on capability gain, breadth, ease of weaponization, and discoverability. NIST CAISI validated the fix. (Anthropic)

  • Anthropic launched Claude Sonnet 5 on June 30, its most agentic Sonnet yet. Now the default model for Free and Pro users. Introductory pricing through August 31 undercuts Sonnet 4.6. (Anthropic)

  • Anthropic's revenue run rate hit $47 billion, nearly double OpenAI's $24-25 billion. Over 1,000 enterprise customers spend $1M+/year. Anthropic expects profitability by 2029, a year ahead of OpenAI. (SaaStr)

  • The United Nations launched the AI for Good Global Commission on July 1, the first UN-level governance body to include CEOs of frontier AI companies. (UN News)

If you haven’t listened to my podcast Mostly Humans: An AI and business podcast for everyone yet, new episodes drop every week!

Episodes can be found below - please like, subscribe, and comment!