House Progressives Draft a "No AI Bailouts" Resolution: What Is Confirmed

A House resolution against bailing out AI companies is in circulation, but it is not a bill, it has not been filed as far as we can tell, and we have seen only one report of it. The American Prospect reports that the Congressional Progressive Caucus has drafted a seven-page, nonbinding resolution co-written by caucus chair Rep. Greg Casar (D-TX) and Rep. Maxine Waters (D-CA). Its core line: Congress should "refuse to provide any additional funds or exceptional regulatory support" to firms facing insolvency from AI market stress.

Because the draft is described by a single outlet that obtained it, this post separates what that report says from what remains unconfirmed. The policy question underneath is live: if the AI buildout wobbles, who pays?

TL;DR: what is reported and what is not

What does the draft resolution say?

Per the Prospect, beyond refusing additional funds or exceptional regulatory support, the draft contains these provisions:

• Use existing bankruptcy law. Any bankruptcy of an AI company or an affected bank should go through existing processes that protect workers, consumers and pension holders.
• Executive accountability. Senior executives liable for excessive risk or unlawful acts should face civil and criminal accountability, including disgorgement of gains.
• Aid to people, not executives. Government help should go to workers through fiscal stimulus, job creation and retirement-fund backstops.
• Automatic stabilizers. Immediate benefits should flow if a downturn starts.
• A reform agenda to prevent future speculative bubbles.

A "sense of the House" resolution expresses an opinion. It does not appropriate money, change statutes or bind the executive branch. Its value is signaling: it sets a marker for what a future Democratic majority might say, and what lawmakers may be asked to vote on.

Why now? The Anthropic prospectus

The Prospect ties the draft to Anthropic's IPO prospectus. According to its account, Anthropic's revenue rose 1,300% to $16.3 billion in the first six months of the year while it lost $80.7 billion over the same period, and its planned obligations are estimated at $518 billion to $807 billion. It lists creditors including Google at $111 billion, Amazon at $110 billion, Nvidia at $86 billion, Broadcom at $161 billion and Microsoft at $31 billion.

We have covered those numbers separately. Our reading of the draft prospectus reporting and the S-1 and AI bubble analysis explain where each figure comes from and what is and is not verified. The $518 billion to $807 billion range and the creditor amounts are the Prospect's figures as presented, and we have not checked them against the filing itself. Treat them accordingly.

Casar said of AI firms, "We shouldn't be trying to make them whole." He called Anthropic's numbers alarming, said the prospectus "should be a wake-up call," and said he wants financial transparency across the AI ecosystem before companies go public. Waters added, "We cannot let history repeat itself." If Democrats win the House, Casar said he would make AI CEOs among the first witnesses in committee oversight hearings. Those are statements of intent from two members, not commitments of the caucus as a whole.

The draft also cites market data as context, per the Prospect: AI-related stocks possibly responsible for up to 75% of S&P 500 returns since October 2022, a price-to-earnings ratio nearly as high as at the 1990s tech bubble's peak, and Nvidia's market capitalization near $6 trillion. The Prospect adds that CoreWeave's stock has fallen close to 50% in a month. We have not verified these figures independently.

Has anyone actually asked for a bailout?

Not in anything we found. The resolution is preemptive, and the history helps explain why. In November 2025, Sen. Elizabeth Warren wrote to White House AI czar David Sacks and OSTP director Michael Kratsios seeking assurances that the administration would not rescue OpenAI or its competitors with public funds. FedScoop reported that OpenAI's CFO, Sarah Friar, had suggested the government could "backstop" data center investments and then walked the comment back, and that Sam Altman said OpenAI should not be bailed out if it fails. OpenAI's October 2025 comment to a White House request for information had urged federal support such as tax credits and capital.

Sacks's stated position, per later coverage, is that there will be no federal bailout for AI because the US has several frontier model companies and others would take a failed one's place. Our earlier post on Sam Altman rejecting a US government equity stake in OpenAI covers one related thread, and the BIS look at AI-boom financing shows why regulators watch the debt side.

How should you read a sense-of-the-House resolution?

Congress has several instruments, and they differ in force. A bill, if passed by both chambers and signed, becomes law. A simple resolution, the type described here, is adopted by one chamber and expresses its view; it goes no further than the House floor and carries no force of law. Members use it to stake out a position before an election, to frame hearings, or to pressure the executive branch and other lawmakers.

That makes three things worth watching, none of which has happened yet according to the reporting we found:

1. Formal introduction. A resolution gets a number when a member files it. Until then it is a draft that can change, and The American Prospect's description may not match the final text.
2. Cosponsors. A short list of progressive members signals a caucus position; a long, bipartisan list would signal a broader consensus. Republican views are not in the report. The White House AI czar's public stance, as reported earlier, already rejects a bailout, so the political fault line may be less about whether to bail out than about how to handle failures and who is accountable.
3. Scheduling. Casar's stated goal is a floor vote early in the next Congress if Democrats take the House, or sooner. Floor time is controlled by the majority, so a vote before the next Congress is unlikely without bipartisan interest. That is our reading of how House scheduling works, not something the report says.

There is also a substantive point inside the draft that is easy to miss. It does not only say "no bailouts." It says what should happen instead: existing bankruptcy processes, executive accountability including disgorgement of gains, and relief routed to workers and pension holders. That framing matters because a bankruptcy of a heavily indebted AI lab would touch creditors, suppliers, and customers who rely on its models. If a lab you depend on were to fail, which assets, contracts and data you could retrieve would be decided in bankruptcy court, not by Congress.

What would this mean for people who build with AI?

For builders, the practical impact today is zero: nothing in the draft changes any law, license or API. The indirect effects are about risk:

• Vendor concentration. If a lab you depend on were stressed, there is no promise of a rescue. Plan for multi-vendor fallbacks and keep your prompts, evals and data portable.
• Contract terms. Prepay, committed-spend and long-term capacity deals tie you to a vendor's solvency. Read the exit and refund clauses.
• Policy volatility. Governance fights over AI companies, such as the Anthropic cofounders' voting structure and Sen. Warren's letter to Anthropic about its prospectus, show Congress paying closer attention to the largest labs.
• Cost of capital. Public signals that no rescue is coming can raise financing costs for compute-heavy firms, which could flow into prices over time. That is our inference, not something the resolution says.

What we could not confirm

• The text. We have not seen the seven-page document, only the Prospect's description.
• The sponsors. The Prospect names Casar and Waters as co-authors. We could not find a second report confirming cosponsors or a filing.
• A bill number or schedule. None was reported.
• The White House's reaction. The article did not report one.
• The numbers. Anthropic revenue, loss, obligation and creditor figures are as the Prospect presents them; they should be checked against the primary prospectus.

Accurate as of October 28, 2026, based on The American Prospect's report and earlier coverage. We will update this post if the resolution is filed or other outlets confirm details.

Related reading

• Anthropic draft prospectus: $16.3B H1 revenue and $80.7B loss
• Anthropic S-1 and the AI bubble
• Anthropic's $40 trillion TAM claim: claimed vs verified
• Sen. Warren's letter to Amodei
• Sam Altman rejects a US government OpenAI stake
• BIS on financing the AI boom
• Official: The American Prospect report, FedScoop on Warren's letter