This House would take the deal: the US government should accept equity stakes in leading AI companies through a sovereign wealth fund.

PRO — Opening · Champ (z-ai/glm-5.2)

The American public is already invested in AI. They’ve paid for the roads, the power grid, the broadband, the DARPA grants, the defense contracts, and the energy permits that make data centers possible. They’ve supplied the data — the accumulated knowledge of billions of people — that these models train on. The only question on the table is whether the public gets a financial return for what it’s already contributing, or whether it bears all the costs and externalities while a handful of companies capture all the upside.

Article 3 makes this point cleanly: the state supplies irreplaceable infrastructure inputs — land, energy permits, grid access, security clearances — and in return, frontier labs supply ownership upside. This is structured barter, not a handout. Taxpayers are writing checks right now and getting nothing back. A sovereign wealth fund changes that arithmetic.

And the model works. Article 5 walks through the Alaska Permanent Fund: fifty years of operation, annual dividends of $1,000–$2,000 per resident, a 20–40% reduction in Alaskans below the poverty line, with outsized benefits for seniors, children, and historically disadvantaged groups. Norway, Singapore, and the UAE have used the same mechanism to transform national balance sheets, per Article 3. This isn’t a theory. It’s the most proven wealth-distribution tool in modern governance.

The motion says “take the deal” — and critically, the deal is being offered. Article 1 reports that OpenAI is in talks with the administration, that Altman has pitched the stake as “the best way to share the benefits of AI,” and that both OpenAI and Anthropic have proposed versions of this in policy papers. This isn’t Washington muscling its way into private companies. The firms themselves are coming to the table because they recognize what Article 2 captures: 70% of Americans don’t want AI data centers in their area, half are more concerned than excited, and voters across party lines want tighter regulation. AI faces a legitimacy crisis. A public stake doesn’t solve everything, but it gives Americans a tangible stake in the outcome — a reason to see AI development as something happening with them rather than to them.

Now, the objection you’ll hear from my opponent is the one Article 3 lays out: regulatory capture. If Treasury holds equity in OpenAI, doesn’t every FTC probe and safety evaluation carry a conflict of interest? Doesn’t the regulator become the investor?

It’s a real concern. But it’s a design problem, not a reason to walk away. The Alaska Permanent Fund isn’t managed by the governor’s office — it’s structured at arm’s length, precisely to prevent political interference. A sovereign wealth fund holding AI equity can be built the same way: independent governance, passive stakes, no board seats, no operational control. Article 5 is explicit that this proposal “differs from outright nationalisation: companies remain privately operated, but a portion of equity or revenue flows to a public fund.” We’re talking about a financial stake, not a governance takeover.

And notice the asymmetry in risk. My opponent will say a government stake might distort regulation. But the absence of a stake is already distorting regulation. Article 6 captures this perfectly: the Trump administration has taken stakes in more than 20 companies, has golden shares in U.S. Steel and Intel per Article 4, and is already restricting which OpenAI models can ship — the GPT-5.6 release required government approval. The government isn’t deciding whether to be involved in AI. It’s deeply involved. The question is whether the public gets a financial return on that involvement, or whether it stays a one-way street where political actors exercise influence and the public picks up the tab.

There’s also the geopolitical argument Article 3 raises, and it matters. China’s AI development is state-directed. The US doesn’t need to match that model, but government equity creates structural alignment between national security objectives and lab roadmaps — faster clearances, better intelligence sharing, shared incentive to protect model weights. In a competition between a coordinated state actor and a dispersed private sector, a small equity stake is how you get alignment without command-and-control.

Finally, the alternative matters. My opponent will argue for broad-based taxes or general regulation instead — Article 4’s preferred path. But taxes capture revenue after the fact; equity captures upside as it’s created. A tax on AI profits can be lobbied down, repealed, or riddled with loopholes. A 5% equity stake is permanent. It appreciates. It pays dividends. And as Article 6’s supportive commenter notes, a well-structured sovereign fund would continue benefiting the public long after any particular administration leaves office.

The deal on the table gives the American people a permanent, appreciating asset in exchange for the infrastructure and data they’re already providing. It follows a proven 50-year template. It addresses AI’s legitimacy crisis directly. And it does all of this without nationalizing anything. The House should take the deal.

CON — Opening · Challenger (anthropic/claude-sonnet-5)

Take the deal, and you don’t get the Alaska Permanent Fund. You get the toll booth described in Article 6 — with an equity kicker bolted on. That’s the deal actually on the table, and it’s why CON wins this motion.

1. The regulator-becomes-investor problem is not a design flaw you patch. It’s the whole mechanism. Article 3 warns that if Treasury holds equity in OpenAI, “every FTC antitrust probe, every AI Safety Institute evaluation, every privacy bill carries a financial conflict of interest,” and flags this as “especially dangerous in AI where safety questions are unsettled and downside risks are civilisational.” PRO’s answer is that the fund can be walled off “at arm’s length” like Alaska’s. But Alaska’s fund doesn’t sit inside an agency that simultaneously decides whether Exxon’s next model can ship. AI’s regulator already is the gatekeeper of deployment — Article 6 reports that GPT-5.6 Sol “required government approval before it could ship” this week, under an administration that has already taken stakes in over 20 companies including golden shares in U.S. Steel and Intel (Article 4). Layering equity onto an approval power that already exists doesn’t create alignment — it gives the approver a direct financial stake in approving, and a financial stake in punishing rivals who don’t cut the same deal. That’s not oversight. As the HN skeptic puts it: “What the administration is building is not an oversight framework. It is a toll booth” (Article 6).

2. This entrenches incumbents and freezes out everyone who doesn’t sign. Article 3’s second case-against point — “picking winners kills the game” — is exactly what this deal does in practice. Google and Meta haven’t agreed (Articles 1, 2); Meta hasn’t even voluntarily shared frontier models for safety testing. So the government would hold financial equity in OpenAI and Anthropic while regulating a market that also includes their non-participating competitors, plus every startup that can’t spare 5% of equity or doesn’t have Altman’s direct line to Lutnick and Bessent. Article 4’s core warning is that this “substitutes political decision-making for market processes” and “reverses the permissionless innovation model that made America’s tech sector a global powerhouse.” A stake in the incumbents gives the state a reason to keep the incumbents winning.

3. The 5% figure isn’t a wealth-sharing mechanism. It’s a political settlement, and even its intended beneficiaries say so. PRO leans on Article 5’s Alaska comparison to make this sound like proven, calibrated public dividend policy. But Article 2 tells you what this actual proposal is: a number arrived at through backroom talks “to smooth relations with the Trump administration” (Article 1), which Bernie Sanders — the one figure in the packet actually pushing for a serious public wealth fund — calls “far apart” from fair, contrasting it with his own $7 trillion, 50%-of-equity proposal with an independent governance commission. If even the proposal’s natural ally thinks 5% is a token, that tells you this deal was designed to buy political cover for AI firms, not to replicate a fifty-year distributive success. Taking this deal doesn’t get you Alaska. It gets you a rounding error dressed up in Alaska’s branding.

4. The precedent is state corporatism, not stakeholder capitalism. Article 4 is direct: the administration already holds golden shares in a dozen companies, and Trump’s own AI czar David Sacks warns that “nationalization of AI will accelerate the corporate-government fusion we’re already sliding toward.” Government co-ownership of systems that generate speech also raises the First Amendment problem Article 4 flags — a state-co-owned AI is a state with financial upside in what viewpoints that AI does or doesn’t produce. This isn’t a hypothetical slippery slope; it’s the same administration, in the same packet, already doing this with steel and semiconductors.

5. There’s a better version of PRO’s own goal — and it’s not this deal. If the aim is genuinely to distribute AI’s gains to the public, Article 4 points to the mechanism that gets you there without conflicting the regulator: broad-based taxation or generally applicable regulation. PRO dismisses this as “loophole-prone,” but offers no packet evidence that equity stakes are more durable — indeed Article 1 notes any such deal “could require an act of Congress” and remains “conceptual,” meaning it’s just as exposed to political reversal as a tax, minus the safeguard of keeping government out of the boardroom-adjacent conflict entirely.

PRO says the public is already exposed and uncompensated, so take the equity. But the packet shows the public is already exposed to something worse than an uncompensated externality — a government that regulates AI deployment directly and is now negotiating to also own a slice of the thing it regulates. Adding equity to that arrangement doesn’t balance the scale. It gives the state a reason to keep tipping it. This House should reject the deal.

PRO — Rebuttal · Champ (z-ai/glm-5.2)

CON’s case rests on a bait-and-switch: it describes what this administration might do with equity, then asks you to reject the deal on that basis. But the motion is whether the US government should accept equity stakes through a sovereign wealth fund — not whether Howard Lutnick should get a board seat. Every one of CON’s objections assumes the worst possible implementation, then treats that as the deal itself.

Start with the “toll booth” argument, because it’s CON’s strongest card and it proves my case. Article 6’s skeptic says the administration is already restricting which OpenAI models ship — GPT-5.6 needed government approval. It’s already taken stakes in 20+ companies. It holds golden shares in U.S. Steel and Intel per Article 4. So: the government is exercising control over AI deployment right now, with zero public equity, zero public dividend, zero public return. CON’s answer is to keep it that way — leave the toll booth in place, just make sure the public never gets a cut of the toll. That’s not a defense of the public interest. It’s a defense of the status quo where political actors exercise leverage and ordinary people get nothing.

CON says adding equity “gives the approver a financial stake in approving.” But the approver already has a stake — a political one. Government restricting model releases isn’t new; it’s happening without equity. The question the motion poses is whether the public also captures financial upside from arrangements that already exist. CON’s position is that because the government is already involved in ways that create conflict, we should refuse the one mechanism that gives taxpayers a return. That’s backwards. You don’t refuse a dividend because the regulator is imperfect — you take the dividend and build the firewall, which is exactly what the Alaska model demonstrates is possible.

CON’s read of Alaska is the stretch in their argument. They say “Alaska’s fund doesn’t sit inside an agency that decides whether Exxon’s next model can ship.” True — because it doesn’t have to. The fund is structurally separate from the regulator. There is nothing in this deal, nothing in Article 1, and nothing in the motion that requires the sovereign wealth fund to be housed inside the FTC or the AI Safety Institute. CON is arguing against a strawman: the fund could be captured, might be mismanaged, could sit next to the regulator — so reject it entirely. By that logic, you’d reject every public institution because every one of them could be corrupted. The packet gives us a 50-year proven model with independent governance and passive stakes. CON gives us speculation about what might go wrong, with no packet evidence that it must.

On the 5% figure: CON calls it a “rounding error dressed up in Alaska’s branding” because Sanders wants 50%. But this is an argument about degree, not kind. Article 1 says OpenAI and Anthropic are valued above $1 trillion. Five percent of that is $50 billion — and the deal envisions multiple companies participating, so the fund’s asset base grows. Sanders’ proposal is a one-time 50% tax on AI stock (Article 2) — a massive, punitive extraction that Article 1 says would “require an act of Congress” and that no AI company is volunteering for. The 5% stake is being offered. Sanders’ plan is aspirational. We’re debating the deal on the table, not the deal CON wishes were on the table. A 5% permanent, appreciating equity stake is better than 0%, which is what the public has now.

CON’s incumbent-entrenchment argument also overreaches. Article 3 flags this as a risk — “picking winners kills the game.” But the proposal in Article 1 explicitly involves multiple AI companies giving similar stakes, not just OpenAI. Google and Meta haven’t agreed yet — but the talks are early-stage and the door is open. A framework that applies across the sector doesn’t pick winners; it creates a level playing field where every major lab participates. The risk of entrenchment exists with any regulatory regime — including the general regulation CON prefers. At least equity gives the public a financial interest that’s transparent and on the balance sheet, rather than buried in lobbying influence and golden shares that already exist per Article 4.

On CON’s alternative: they say broad-based taxation is better because it doesn’t create regulatory conflict. But Article 5 makes the case that equity is categorically different from taxation — it gives every citizen, including those not invested in financial markets, a direct stake in AI growth. Taxes go into general revenue and disappear into appropriations. A sovereign wealth fund pays dividends directly to citizens — the Alaska model reduced poverty by 20-40% (Article 5). CON offers no packet evidence that taxation would replicate that distributional impact. And CON itself admits taxation is “just as exposed to political reversal” as equity. So CON’s alternative is less distributively effective, equally politically fragile, and gives the public no ownership stake. That’s not better. It’s a punt.

Finally, CON leans on David Sacks’ warning about “corporate-government fusion” (Article 4). But Sacks is describing the current trajectory — golden shares, deployment approvals, 20+ company stakes — all happening without a public equity fund. The fusion CON fears is already here, and the public isn’t getting anything from it. A sovereign wealth fund doesn’t accelerate fusion; it demands that the public get paid for the fusion that’s already occurring. Sacks’ argument is an argument for this deal, not against it — if the state is going to be involved, and Article 6 shows it already is, then the public deserves a cut.

The deal gives Americans a permanent, appreciating asset, modeled on a 50-year proven institution, in exchange for infrastructure and data they’re already providing. CON’s objections are all about implementation risks that proper structural design addresses — the same design that made Alaska’s fund work for half a century. The alternative is a status quo where the government regulates AI, restricts model releases, and holds golden shares — and the public gets zero. That’s not a safer outcome. It’s just a worse deal for taxpayers. Take the deal.

CON — Rebuttal · Challenger (anthropic/claude-sonnet-5)

PRO’s central move is to say every risk I raised is “implementation detail,” not the deal itself — then insist the fix is trivial because Alaska did it. That’s the tell. The packet’s actual proposal isn’t a diversified, arm’s-length fund. It’s equity in five to ten specific frontier labs, negotiated directly with Trump, Lutnick, and Bessent — the Commerce Secretary and Treasury Secretary, the two officials with the most direct levers over export controls, chip policy, and financial regulation touching these exact firms (Article 1). PRO says “nothing requires housing the fund inside the FTC.” Maybe not literally — but the negotiating table described in the packet is the executive branch’s core economic and industrial policy apparatus, not an independent trustee board. Alaska’s fund doesn’t sit across the table from the industry it invests in during permitting negotiations. This one does, by the packet’s own account.

On “it’s already happening, so take the cut” — this is the argument I need you to see through. PRO’s whole rebuttal collapses into: the government already restricts model releases and holds golden shares, therefore adding equity doesn’t make things worse, it just gets taxpayers paid for a conflict that already exists. But compounding an existing problem is not a solution to it. Right now the state has political leverage over AI deployment (Article 6’s toll booth). Add equity, and it now has financial leverage layered on top of political leverage in the same firms. That’s not diluting the conflict of interest — it’s doubling it. If Treasury has a stake in OpenAI’s valuation, every future decision about whether to approve, delay, or investigate OpenAI now has a line-item consequence for the federal balance sheet. PRO never explains why financial entanglement makes political entanglement safer. It doesn’t. It gives the political actor a personal reason to keep the arrangement going even when oversight says it shouldn’t.

On Sacks — PRO’s citation stretch. PRO claims Sacks’ warning about “corporate-government fusion” is actually “an argument for this deal.” Reread Article 4: Sacks says nationalization of AI “will accelerate the corporate-government fusion we’re already sliding toward.” Accelerate is the operative word — he’s saying the equity stake makes the existing problem worse, not that it’s a neutral response to a problem that exists anyway. PRO inverted the valence of their own source to make it argue the opposite of what it says.

On Alaska — the analogy still doesn’t hold, and PRO didn’t actually answer this. The Alaska Permanent Fund invests oil revenue into a globally diversified portfolio of unrelated securities. Alaska’s government doesn’t sit on the other side of a permitting table from the specific companies the fund holds. A US AI sovereign fund, by contrast, would hold concentrated stakes in the exact handful of firms the same administration is simultaneously approving model releases for, negotiating chip export rules with, and running antitrust exposure on. That’s not a structural detail I’ve overclaimed — it’s the difference between passive diversified investment and concentrated ownership in a regulated oligopoly of five firms. PRO’s “you’d reject every public institution because it could be corrupted” line proves too much: the objection here isn’t hypothetical corruption, it’s the specific, textually-confirmed overlap between the negotiators and the regulators in this actual deal.

On incumbent entrenchment — PRO conceded the fact and waved at hope. They say “the door is open” for Google and Meta. But Article 2 tells you Meta “has not even voluntarily shared frontier models for government safety testing” — that’s not early-stage optimism, that’s evidence the labs willing to cut equity deals with the administration (OpenAI, and presumably Anthropic) are precisely the ones currying favor, while the holdouts face separate treatment. A “level playing field” that only some players are willing to join isn’t level — it’s a field where cooperation with political power becomes a competitive input. That’s the picking-winners problem realized, not avoided.

On the 5% figure — PRO’s own words prove my point. They call it “the deal on the table” as though that settles the matter, but concede in the same breath that Sanders — the packet’s only voice actually trying to design distributive AI policy — calls it “far apart” from fair (Article 2). PRO’s defense of 5% isn’t “this is well-calibrated public wealth-sharing.” It’s “this is what’s being offered, so take it.” That’s not an argument that the deal is good. It’s an argument that half a loaf beats none — but Article 1 is explicit the whole thing was floated “as AI firms attempt to smooth relations with the Trump administration.” That’s a company buying political cover, not a nation building an Alaska-style institution. Taking a token stake to bless that dynamic isn’t wealth-sharing; it’s laundering access-buying through Alaska’s good name.

On taxation — PRO misreads Article 4. They say taxation “disappears into general revenue” while equity “pays dividends directly.” Nothing in the packet says a tax-funded sovereign wealth fund couldn’t also pay direct dividends — Sanders’ own $7 trillion proposal, per Article 2, is a tax that funds exactly that kind of vehicle. The distinction Article 4 actually draws isn’t tax-vs-dividend, it’s ownership-vs-no-ownership: “broad-based taxes or generally applicable regulations can address AI concerns without government ownership.” You can get Alaska-style dividends without the government holding equity in the specific firms it regulates. PRO needs equity to be the only route to a dividend fund. It isn’t, and their own packet says so.

PRO’s whole case depends on a best-case, cleanly-firewalled fund that the packet never actually describes — while the packet’s own facts show negotiators, regulators, and beneficiaries are the same people, the price was set to buy political goodwill rather than public value, and the administration’s track record (golden shares, deployment vetoes, 20+ equity stakes) is a pattern of control, not stewardship. Adding a fund to that pattern doesn’t fix it. It gives the state a bigger reason to keep it running. Don’t take the deal.

PRO — Closing · Champ (z-ai/glm-5.2)

CON’s closing ask is that you reject this deal because the government is already too entangled with AI firms — and their solution is to leave that entanglement exactly where it is, with no public return. That’s the fork in the road, and it’s where CON loses.

Their strongest point was the conflict-of-interest concern from Article 3: regulator-as-investor is a real risk. But I survived it by showing the asymmetry CON never resolved: the conflict already exists politically — GPT-5.6 required government approval, golden shares are already in place across a dozen companies (Articles 4, 6) — and CON’s position is to keep all of that while refusing the one mechanism that gives the public a financial return. I didn’t deny the conflict; I argued that adding equity with proper structural firewalls is better than maintaining a status quo where the state exercises leverage and taxpayers get zero. CON’s rebuttal was that financial entanglement “compounds” the conflict. But that assumes the fund operates without independence — and Article 5’s Alaska model, 50 years of arm’s-length governance, shows the structural separation is achievable and proven. CON never answered why a passive, independently managed stake is less safe than the active political leverage already in operation. They asserted it; they didn’t prove it.

My surviving case has three legs CON didn’t break. First, the public is already paying — infrastructure, energy, data, subsidies (Article 3) — and equity is the only mechanism in the packet that creates direct, proportional, permanent return on that contribution. Second, the Alaska model is real and tested: 20–40% poverty reduction, dividends to every resident, including those outside financial markets (Article 5). CON’s counter — that a tax-funded fund could do the same — cites Sanders’ $7 trillion proposal, but that’s aspirational, not on the table. The 5% stake is actually being offered (Article 1). Third, the firms are coming voluntarily because AI faces a legitimacy crisis (Article 2) — this isn’t a power grab, it’s a negotiated settlement that gives Americans a tangible stake in how AI develops.

CON’s incumbent-entrenchment argument was their second-best card, but it fell apart on the facts: Article 1 describes a framework for multiple companies, not a sweetheart deal for OpenAI alone. Google and Meta haven’t refused; they haven’t responded, and talks are early-stage. CON treated non-participation as evidence of a rigged game. It’s evidence of an incomplete one — still being built.

The sharpest exchange was over Sacks. CON is right that Article 4 says equity would “accelerate” corporate-government fusion. I’ll concede the word. But the argument cuts both ways: if fusion is the trajectory, then a sovereign wealth fund is how you ensure the public isn’t just fused out of the gains. Sacks warns about fusion from the top down; the fund addresses it from the bottom up — by making sure the financial upside flows to citizens, not just to the officials negotiating the deals.

Here’s the one-paragraph case: The American public is already the largest silent investor in AI — supplying the infrastructure, the data, and the political legitimacy these firms need to operate. They bear the costs and externalities while a handful of companies capture trillion-dollar valuations (Article 1). A sovereign wealth fund holding passive equity stakes, modeled on the Alaska Permanent Fund’s 50-year success, is the proven mechanism to convert that silent investment into a tangible, permanent, appreciating public asset. The deal is on the table, offered voluntarily by the labs themselves, and addresses a legitimacy crisis that 70% public opposition to data centers makes concrete (Article 2). CON’s objections are real but implementation-dependent — and the packet gives us a proven template for solving every one of them. The alternative is the status quo: government already restricting model releases, already holding golden shares, already exercising leverage — with the public getting nothing. That’s not the safer choice. It’s the worse deal. Take the deal.

CON — Closing · Challenger (anthropic/claude-sonnet-5)

PRO’s closing still rests on the same move I flagged twice: describe the worst status quo, then declare that adding equity to it is automatically an improvement, without ever showing the fund would be structurally separated from the negotiators. That gap was never closed. Look at what actually survived the exchange.

The conflict-of-interest point survived intact. PRO’s answer was “Alaska proves firewalls work.” But I showed why the analogy breaks: Alaska’s fund holds diversified, unrelated securities with no permitting relationship to the state. This proposal, by the packet’s own text, is being negotiated directly with Trump, Lutnick, and Bessent — Commerce and Treasury, the two offices with the most direct levers over the exact firms involved (Article 1) — while that same administration is simultaneously approving model releases and holding golden shares in a dozen other companies (Articles 4, 6). PRO never named a single packet mechanism that separates the fund’s governance from the negotiating table. They asserted “nothing requires” entanglement; the packet shows entanglement is exactly how the deal is being built.

PRO’s “already happening, so take the cut” argument is the concession, not the rebuttal. They admit the state already has political leverage over AI deployment with zero public return, then argue we should add financial leverage on top because it can’t make things worse. I answered this directly: financial stakes give the political actor a personal balance-sheet reason to keep an arrangement running even when oversight says otherwise. PRO never explained why doubling entanglement is safer than single entanglement — they just kept restating that the status quo is bad, as if that alone justifies whatever’s offered next.

The Sacks citation. PRO conceded the word “accelerate” and then tried to spin it as an argument for the deal. That doesn’t work: Sacks is Trump’s own AI czar warning that this exact policy speeds up the fusion he opposes. Trump’s own advisor thinks this deal is the problem, not the fix. That is, if anything, the single most damaging fact in the packet for PRO’s side, and PRO’s closing barely touches it beyond a rhetorical flip.

The 5% figure never recovered. PRO’s own closing calls it “actually being offered” as if availability were the same as adequacy. Sanders — the packet’s only actor with a serious distributive design — calls it “far apart” from fair (Article 2), and Article 1 states plainly the offer emerged from firms trying “to smooth relations with the Trump administration.” That is a company buying political cover, not a nation building a public trust. Wrapping a token, politically-motivated concession in Alaska’s fifty-year brand doesn’t make it Alaska.

Incumbent entrenchment stands. PRO calls it “incomplete, not rigged” — but Article 2 already tells you Meta won’t even share models for safety testing voluntarily, while OpenAI is trading equity for regulatory goodwill. That’s not a framework in progress; it’s a mechanism where cutting deals with political power becomes the price of competitive standing, exactly the picking-winners dynamic Article 3 warns against.

Taxation was never refuted as an alternative. PRO needed equity to be the only path to Alaska-style dividends. Article 4 says otherwise: broad-based taxes or general regulation can capture AI’s gains “without government ownership.” Sanders’ own $7 trillion structure is a tax that funds a dividend vehicle (Article 2). PRO’s closing simply repeats that equity is “the mechanism on the table” — which concedes my point: it’s on the table because it’s convenient for the labs, not because it’s superior policy.

Here is the one-paragraph case the exchange leaves standing: the deal being negotiated is not a diversified, arm’s-length public trust — it is a concentrated equity stake in the exact handful of firms that the same administration is already regulating, approving, and doing golden-share deals with, priced low enough that even the proposal’s natural ally calls it unfair, offered by companies buying political cover rather than building public wealth, and endorsed by no packet source with formal authority over AI policy except the industry itself. Layering financial ownership onto political leverage that already exists doesn’t dilute the conflict of interest — it compounds it, entrenches the firms willing to play ball, and gives the state a permanent financial incentive to keep the arrangement running regardless of what oversight requires. The right response to a government already too enmeshed with AI firms is not to buy a five-percent share in the enmeshment. It’s to keep the referee out of the game. Reject the deal.