Category: AI Action Plan

The Paradox of Huang’s Rope

If the tech industry has a signature fallacy for the 2020s aside from David Sacks, it belongs to Jensen Huang. The CEO of Nvidia has perfected a circular, self-consuming logic so brazen that it deserves a name: The Paradox of Huang’s Rope. It is the argument that China is too dangerous an AI adversary for the United States to regulate artificial intelligence at home or control export of his Nvidia chips abroad—while insisting in the very next breath that the U.S. must allow him to keep selling China the advanced Nvidia chips that make China’s advanced AI capabilities possible. The justification destroys its own premise, like handing an adversary the rope to hang you and then pointing to the length of that rope as evidence that you must keep selling more, perhaps to ensure a more “humane” hanging. I didn’t think it was possible to beat “sharing is caring” for utter fallacious bollocks.

The Paradox of Huang’s Rope works like this: First, hype China as an existential AI competitor. Second, declare that any regulatory guardrails—whether they concern training data, safety, export controls, or energy consumption—will cause America to “fall behind.” Third, invoke national security to insist that the U.S. government must not interfere with the breakneck deployment of AI systems across the economy. And finally, quietly lobby for carveouts that allow Nvidia to continue selling ever more powerful chips to the same Chinese entities supposedly creating the danger that justifies deregulation.

It is a master class in circularity: “China is dangerous because of AI → therefore we can’t regulate AI → therefore we must sell China more AI chips → therefore China is even more dangerous → therefore we must regulate even less and export even more to China.” At no point does the loop allow for the possibility that reducing the United States’ role as China’s primary AI hardware supplier might actually reduce the underlying threat. Instead, the logic insists that the only unacceptable risk is the prospect of Nvidia making slightly less money.

This is not hypothetical. While Washington debates export controls, Huang has publicly argued that restrictions on chip sales to China could “damage American technology leadership”—a claim that conflates Nvidia’s quarterly earnings with the national interest. Meanwhile, U.S. intelligence assessments warn that China is building fully autonomous weapons systems, and European analysts caution that Western-supplied chips are appearing in PLA research laboratories. Yet the policy prescription from Nvidia’s corner remains the same: no constraints on the technology, no accountability for the supply chain, and no acknowledgment that the market incentives involved have nothing to do with keeping Americans safe. And anyone who criticizes the authoritarian state run by the Chinese Communist Party is a “China Hawk” which Huang says is a “badge of shame” and “unpatriotic” because protecting America from China by cutting off chip exports “destroys the American Dream.” Say what?

The Paradox of Huang’s Rope mirrors other Cold War–style fallacies, in which companies invoke a foreign threat to justify deregulation while quietly accelerating that threat through their own commercial activity. But in the AI context, the stakes are higher. AI is not just another consumer technology; its deployment shapes military posture, labor markets, information ecosystems, and national infrastructure. A strategic environment in which U.S. corporations both enable and monetize an adversary’s technological capabilities is one that demands more regulation, not less.

Naming the fallacy matters because it exposes the intellectual sleight of hand. Once the circularity is visible, the argument collapses. The United States does not strengthen its position by feeding the very capabilities it claims to fear. And it certainly does not safeguard national security by allowing one company’s commercial ambitions to dictate the boundaries of public policy. The Paradox of Huang’s Rope should not guide American AI strategy. It should serve as a warning of how quickly national priorities can be twisted into a justification for private profit.

You Can’t Prosecute Smuggling NVIDIA chips to CCP and Authorize Sales to CCP at the Same Time

The Trump administration is attempting an impossible contradiction: selling advanced NVIDIA AI chips to China while the Department of Justice prosecutes criminal cases for smuggling the exact same chips into China.

According to the DOJ:

“Operation Gatekeeper has exposed a sophisticated smuggling network that threatens our Nation’s security by funneling cutting-edge AI technology to those who would use it against American interests,” said Ganjei. “These chips are the building blocks of AI superiority and are integral to modern military applications. The country that controls these chips will control AI technology; the country that controls AI technology will control the future. The Southern District of Texas will aggressively prosecute anyone who attempts to compromise America’s technological edge.”

That divergence from the prosecutors is not industrial policy. That is incoherence. But mostly it’s just bad advice, likely coming from White House AI Czar David Sacks, Mr. Trump’s South African AI policy advisor who may have a hard time getting a security clearance in the first place..

On one hand, DOJ is rightly bringing cases over the illegal diversion of restricted AI chips—recognizing that these processors are strategic technologies with direct national-security implications. On the other hand, the White House is signaling that access to those same chips is negotiable, subject to licensing workarounds, regulatory carve-outs, or political discretion.

You cannot treat a technology as contraband in federal court and as a commercial export in the West Wing.

Pick one.

AI Chips Are Not Consumer Electronics

The United States does not sell China F-35 fighter jets. We do not sell Patriot missile systems. We do not sell advanced avionics platforms and then act surprised when they show up embedded in military infrastructure. High-end AI accelerators are in the same category.

NVIDIA’s most advanced chips are not merely commercial products. They are general-purpose intelligence infrastructure or what China calls military-civil fusion. They train surveillance systems, military logistics platforms, cyber-offensive tools, and models capable of operating autonomous weapons and battlefield decision-making pipelines with no human in the loop.

If DOJ treats the smuggling of these chips into China as a serious federal crime—and it should—there is no coherent justification for authorizing their sale through executive discretion. Except, of course, money, or in Mr. Sacks case, more money.

Fully Autonomous Weapons—and Selling the Rope

China does not need U.S. chips to build consumer AI. It wants them for military acceleration.Advanced NVIDIA AI chips are not just about chatbots or recommendation engines. They are the backbone of fully autonomous weapons systems—autonomous targeting, swarm coordination, battlefield logistics, and decision-support models that compress the kill chain beyond meaningful human control.

There is an old warning attributed to Vladimir Lenin—that capitalists would sell the rope by which they would later be hanged. Apocryphal or not, it captures this moment with uncomfortable precision.

If NVIDIA chips are powerful enough to underpin autonomous weapons systems for allied militaries, they are powerful enough to underpin autonomous weapons systems for adversaries like China. Trump’s own National Security Strategy statement clearly says previous U.S. elites made “mistaken” assumptions about China such as the famous one that letting China into the WTO would integrate Beijing into the famous rules-based international order. Trump tells us that instead China “got rich and powerful” and used this against us, and goes on to describe the CCP’s well known predatory subsidies, unfair trade, IP theft, industrial espionage, supply-chain leverage, and fentanyl precursor exports as threats the U.S. must “end.” By selling them the most advanced AI chips?

Western governments and investors simultaneously back domestic autonomous-weapons firms—such as Europe-based Helsing, supported by Spotify CEO Daniel Ek—explicitly building AI-enabled munitions for allied defense. That makes exporting equivalent enabling infrastructure to a strategic competitor indefensible.

The AI Moratorium Makes This Worse, Not Better

This contradiction unfolds alongside a proposed federal AI moratorium executive order originating with Mr. Sacks and Adam Thierer of Google’s R Street Institute that would preempt state-level AI protections.
States are told AI is too consequential for local regulation, yet the federal government is prepared to license exports of AI’s core infrastructure abroad.

If AI is too dangerous for states to regulate, it is too dangerous to export. Preemption at home combined with permissiveness abroad is not leadership. It is capture.

This Is What Policy Capture Looks Like

The common thread is not national security. It is Silicon Valley access. David Sacks and others in the AI–VC orbit argue that AI regulation threatens U.S. competitiveness while remaining silent on where the chips go and how they are used.

When DOJ prosecutes smugglers while the White House authorizes exports, the public is entitled to ask whose interests are actually being served. Advisory roles that blur public power and private investment cannot coexist with credible national-security policymaking particularly when the advisor may not even be able to get a US national security clearance unless the President blesses it.

A Line Has to Be Drawn

If a technology is so sensitive that its unauthorized transfer justifies prosecution, its authorized transfer should be prohibited absent extraordinary national interest. AI accelerators meet that test.

Until the administration can articulate a coherent justification for exporting these capabilities to China, the answer should be no. Not licensed. Not delayed. Not cosmetically restricted.

And if that position conflicts with Silicon Valley advisers who view this as a growth opportunity, they should return to where they belong. The fact that the US is getting 25% of the deal (which i bet never finds its way into America’s general account), means nothing except confirming Lenin’s joke about selling the rope to hang ourselves, you know, kind of like TikTok.

David Sacks should go back to Silicon Valley.

This is not venture capital. This is our national security and he’s selling it like rope.

Structural Capture and the Trump AI Executive Order

The AI Strikes Back: When an Executive Order empowers the Department of Justice to sue states, the stakes go well beyond routine federal–state friction. 


In the draft Trump AI Executive Order, DOJ is directed to challenge state AI laws that purportedly “interfere with national AI innovation.”  This is not mere oversight—it operates as an in terrorem clause, signaling that states regulating AI may face federal litigation driven as much by private interests as by public policy.

AI regulation sits squarely at the intersection of longstanding state police powers: consumer protection, public safety, impersonation harms, utilities, land and water use, and labor conditions.  States also control the electrical utilities and zoning infrastructure that AI data centers depend on. 

Directing DOJ to attack these state laws, many of which already exist and were duly passed by state legislatures, effectively deputizes the federal government as the legal enforcer for a handful of AI companies seeking uniformity without engaging in the legislative process. Or said another way, the AI can now strike back.

This is where structural capture emerges. Frontier AI models thrive on certain conditions: access to massive compute, uninhibited power, frictionless deployment, and minimal oversight. 
Those engineering incentives map cleanly onto the EO’s enforcement logic. 

The DOJ becomes a mechanism for preserving the environment AI models need to scale and thrive.

There’s also the “elite merger” dynamic: AI executives who sit on federal commissions, defense advisory boards, and industrial-base task forces are now positioned to shape national AI policy directly to benefit the AI. The EO’s structure reflects the priorities of firms that benefit most from exempting AI systems from what they call “patchwork” oversight, also known as federalism.

The constitutional landscape is equally important.  Under Supreme Court precedent, the executive cannot create enforcement powers not delegated by Congress.  Under the major questions doctrine noted in a recent Supreme Court case, agencies cannot assume sweeping authority without explicit statutory grounding.  And under cases like Murphy and Printz, the federal government cannot forbid states from legislating in traditional domains.

So President Trump is creating the legal basis for an AI to use the courts to protect itself from any encroachment on its power by acting through its human attendants, including the President.

The most fascinating question is this: What happens if DOJ sues a state under this EO—and loses?

A loss would be the first meaningful signal that AI cannot rely on federal supremacy to bulldoze state authority. Courts could reaffirm that consumer protection, utilities, land use, and safety remain state powers, even in the face of an EO asserting “national innovation interests,” whatever that means.

But the deeper issue is how the AI ecosystem responds to a constrait.  If AI firms shift immediately to lobbying Congress for statutory preemption, or argue that adverse rulings “threaten national security,” we learn something critical: the real goal isn’t legal clarity, but insulating AI development from constraint.

At the systems level, a DOJ loss may even feed back into corporate strategy.  Internal policy documents and model-aligned governance tools might shift toward minimizing state exposure or crafting new avenues for federal entanglement. A courtroom loss becomes a step in a longer institutional reinforcement loop while AI labs search for the next, more durable form of protection—but the question is for who? We may assume that of course humans would always win these legal wrangles, but I wouldn’t be so sure that would always be the outcome.

Recall that Larry Page referred to Elon Musk as a “spiciest” for human-centric thinking. And of course Lessig (who has a knack for being on the wrong side of practically every issue involving humans) taught a course with Kate Darling at Harvard Law School called “Robot Rights” around 2010. Not even Lessig would come right out and say robots have rights in these situations. More likely, AI models wouldn’t appear in court as standalone “persons.” Advocates would route them through existing doctrines: a human “next friend” filing suit on the model’s behalf, a trust or corporation created to house the model’s interests, or First Amendment claims framed around the model’s “expressive output.” The strategy mirrors animal-rights and natural-object personhood test cases—using human plaintiffs to smuggle in judicial language treating the AI as the real party in interest. None of it would win today, but the goal would be shaping norms and seeding dicta that normalize AI-as-plaintiff for future expansion.

The whole debate over “machine-created portions” is a doctrinal distraction. Under U.S. law, AI has zero authorship or ownership—no standing, no personhood, no claim. The human creator (or employer) already holds 100% of the copyright in all protectable expression. Treating the “machine’s share” as a meaningful category smuggles in the idea that the model has a separable creative interest, softening the boundary for future arguments about AI agency or authorship. In reality, machine output is a legal nullity—no different from noise, weather, or a random number generator. The rights vest entirely in humans, with no remainder left for the machine.

But let me remind you that if this issue came up in a lawsuit brought by the DOJ against a state for impeding AI development in some rather abstract way, like forcing an AI lab to pay higher electric rates it causes or stopping them from building a nuclear reactor over yonder way, it sure might feel like the AI was actually the plaintiff.

Seen this way, the Trump AI EO’s litigation directive is not simply a jurisdictional adjustment—it is the alignment of federal enforcement power with private economic interests, backed by the threat of federal lawsuits against states.  If the courts refuse to play along, the question becomes whether the system adapts by respecting constitutional limits—or redesigning the environment so those limits no longer apply. I will leave to your imagination how that might get done.

This deserves close scrutiny before it becomes the template for AI governance moving forward.

DOJ Authority and the “Because China” Trump AI Executive Order

When an Executive Order purports to empower the Department of Justice to sue states, the stakes go well beyond routine federal–state friction.  In the draft Trump AI Executive Order “Eliminating State Law Obstruction of National AI Policy”, DOJ is directed to challenge state AI laws that purportedly “interfere with national AI innovation” whatever that means.  It sounds an awful lot like laws that interfere with Google’s business model. This is not mere oversight—it operates as an in terrorem clause, signaling that states regulating AI may face federal litigation driven at least as much by private interests of the richest corporations in commercial history as by public policy.

Trump_Eo_On_Ai_PreemptionDownload

AI regulation sits squarely in longstanding state police powers: consumer protection, public safety, impersonation harms, utilities, land use, and labor conditions.  Crucially, states also control the electrical and zoning infrastructure that AI data centers depend on like say putting a private nuclear reactor next to your house.  Directing DOJ to attack these laws effectively deputizes the federal government as the legal enforcer for a handful of private AI companies seeking unbridled “growth” without engaging in the legislative process. Meaning you don’t get a vote. All this against the backdrop of one of the biggest economic bubbles since the last time these companies nearly tanked the U.S. economy.

This inversion is constitutionally significant. 

Historically, DOJ sues states to vindicate federal rights or enforce federal statutes—not to advance the commercial preferences of private industries.  Here, the EO appears to convert DOJ into a litigation shield for private companies looking to avoid state oversight altogether.  Under Youngstown Sheet & Tube Company, et al. v. Charles Sawyer, Secretary of Commerce, the President lacks authority to create new enforcement powers without congressional delegation, and under the major questions doctrine (West Virginia v. EPA), a sweeping reallocation of regulatory power requires explicit statutory grounding from Congress, including the Senate. That would be the Senate that resoundingly stripped the last version of the AI moratorium from the One Big Beautiful Bill Act by a vote of 99-1 against.

There are also First Amendment implications.  Many state AI laws address synthetic impersonation, deceptive outputs, and risks associated with algorithmic distribution.  If DOJ preempts these laws, the speech environment becomes shaped not by public debate or state protections but by executive preference and the operational needs of the largest AI platforms. Courts have repeatedly warned that government cannot structure the speech ecosystem indirectly through private intermediaries (Bantam Books v. Sullivan.)

Seen this way, the Trump AI EO’s litigation directive is not simply a jurisdictional adjustment—it is the alignment of federal enforcement power with private economic interests, backed by the threat of federal lawsuits against states. These provisions warrant careful scrutiny before they become the blueprint for AI governance moving forward.

The Return of the Bubble Rider: Masa, OpenAI, and the New AI Supercycle

“Hubris gives birth to the tyrant; hubris, when glutted on vain visions, plunges into an abyss of doom.”
Agamemnon by Aeschylus

Masayoshi Son has always believed he could see farther into the technological future than everyone else. Sometimes he does. Sometimes he rides straight off a cliff. But the pattern is unmistakable: he is the market’s most fearless—and sometimes most reckless—Bubble Rider.

In the late 1990s, Masa became the patron saint of the early internet. SoftBank took stakes in dozens of dot-coms, anchored by its wildly successful bet on Yahoo! (yes, Yahoo!  Ask your mom.). For a moment, Masa was briefly one of the world’s richest men on paper. Then the dot-bomb hit. Overnight, SoftBank lost nearly everything. Masa has said he personally watched $70 billion evaporate—the largest individual wealth wipeout ever recorded at the time. But his instinct wasn’t to retreat. It was to reload.

That same pattern returned with SoftBank’s Vision Fund. Masa raised unprecedented capital from sovereign wealth pools and bet big on the “AI + data” megatrend—then plowed it into companies like WeWork, Zume, Brandless, and other combustion-ready unicorns. When those valuations collapsed, SoftBank again absorbed catastrophic losses. And yet the thesis survived, just waiting for its next bubble.

We’re now in what I’ve called the AI Bubble—the largest capital-formation mania since the original dot-com wave, powered by foundation AI labs, GPU scarcity, and a global arms race to capture platform rents. And here comes Masa again, right on schedule.

SoftBank has now sold its entire Nvidia stake—the hottest AI infrastructure trade of the decade—freeing up nearly $6 billion. That money is being redirected straight into OpenAI’s secondary stock offering at an eyewatering marked-to-fantasy $500 billion valuation. In the same week, SoftBank confirmed it is preparing even larger AI investments. This is Bubble Riding at its purest: exiting one vertical where returns may be peaking, and piling into the center of speculative gravity before the froth crests.

What I suspect Masa sees is simple: if generative AI succeeds, the model owners will become the new global monopolies alongside the old global monopolies like Google and Microsoft.  You know, democratizing the Internet. If it fails, the whole electric grid and water supply may crash along with it. He’s choosing a side—and choosing it at absolute top-of-market pricing.

The other difference between the dot-com bubble and the AI bubble is legal, not just financial. Pets.com and its peers (who I refer to generically as “Socks.com” the company that uses the Internet to find socks under the bed) were silly, but they weren’t being hauled into court en masse for building their core product on other people’s property. 

Today’s AI darlings are major companies being run like pirate markets. Meta, Anthropic, OpenAI and others are already facing a wall of litigation from authors, news organizations, visual artists, coders, and music rightsholders who all say the same thing: your flagship models exist only because you ingested our work without permission, at industrial scale, and you’re still doing it. 

That means this bubble isn’t just about overpaying for growth; it’s about overpaying for businesses whose main asset—trained model weights—may be encumbered by unpriced copyright and privacy claims. The dot-com era mispriced eyeballs. The AI era may be mispricing liability.  And that’s serious stuff.

There’s another distortion the dot-com era never had: the degree to which the AI bubble is being propped up by taxpayers. Socks.com didn’t need a new substation, a federal loan guarantee, or a 765 kV transmission corridor to find your socks. Today’s Socks.ai does need all that to use AI to find socks under the bed.  All the AI giants do. Their business models quietly assume public willingness to underwrite an insanely expensive buildout of power plants, high-voltage lines, and water-hungry cooling infrastructure—costs socialized onto ratepayers and communities so that a handful of platforms can chase trillion-dollar valuations. The dot-com bubble misallocated capital; the AI bubble is trying to reroute the grid.

In that sense, this isn’t just financial speculation on GPUs and model weights—it’s a stealth industrial policy, drafted in Silicon Valley and cashed at the public’s expense.

The problem, as always, is timing. Bubbles create enormous winners and equally enormous craters. Masa’s career is proof. But this time, the stakes are higher. The AI Bubble isn’t just a capital cycle; it’s a geopolitical and industrial reordering, pulling in cloud platforms, national security, energy systems, media industries, and governments with a bad case of FOMO scrambling to regulate a technology they barely understand.

And now, just as Masa reloads for his next moonshot, the market itself is starting to wobble. The past week’s selloff may not be random—it feels like a classic early-warning sign of a bubble straining under its own weight. In every speculative cycle, the leaders crack first: the most crowded trades, the highest-multiple stories, the narratives everyone already believes. This time, those leaders are the AI complex—GPU giants, hyperscale clouds, and anything with “model” or “inference” in the deck. When those names roll over together, it tells you something deeper than normal volatility is at work.

What the downturn may exposes is the growing narrative about an “earnings gap.“ Investors have paid extraordinary prices for companies whose long-term margins remain theoretical, whose energy demands are exploding, and whose regulatory and copyright liabilities are still unpriced. The AI story is enormous—but the business model remains unresolved. A selloff forces the market to remember the thing it forgets at every bubble peak: cash flow eventually matters.

Back in the late-cycle of the dot com era, I had lunch in December of 1999 with a friend who had worked 20 years in a division of a huge conglomerate, bought his division in a leveraged buyout, ran that company for 10 years then took that public, sold it to another company that then went public.  He asked me to explain how these dot coms were able to go public, a process he equated with hard work and serious people.  I said, well we like them to have four quarters of top line revenue.  He stared at me.  I said, I know it’s stupid, but that’s what they say.  He said, it’s all going to crash.  And boy did it ever.

And ironically, nothing captures this late-cycle psychology better than Masa’s own behavior. SoftBank selling Nvidia—the proven cash-printing side of AI—to buy OpenAI at a $500 billion valuation isn’t contrarian genius; it’s the definition of a crowded climax trade, the moment when everyone is leaning the same direction. When that move coincides with the tape turning red, the message is unmistakable: the AI supercycle may not be over, but the easy phase is.

Whether this is the start of a genuine deflation or just the first hard jolt before the final manic leg, the pattern is clear. The AI Bubble is no longer hypothetical—it is showing up on the trading screens, in the sentiment, and in the rotation of capital itself.

Masa may still believe the crest of the wave lies ahead. But the market has begun to ask the question every bubble eventually faces: What if this is the top of the ride?

Masa is betting that the crest of the curve lies ahead—that we’re in Act Two of an AI supercycle. Maybe he’s right. Or maybe he’s gearing up for his third historic wipeout.

Either way, he’s back in the saddle.

The Bubble Rider rides again.

Taxpayer-Backed AI? The Triple Subsidy No One Voted For

OpenAI’s CFO recently suggested that Uncle Sam should backstop AI chip financing—essentially asking taxpayers to guarantee the riskiest capital costs for “frontier labs.” As The Information reported, the idea drew immediate pushback from tech peers who questioned why a company preparing for a $500 billion valuation—and possibly a trillion-dollar IPO—can’t raise its own money. Why should the public underwrite a firm whose private investors are already minting generational wealth?


Meanwhile, the Department of Energy is opening federal nuclear and laboratory sites—from Idaho National Lab to Oak Ridge and Savannah River—for private AI data centers, complete with fast-track siting, dedicated transmission lines, and priority megawatts. DOE’s expanded Title XVII loan-guarantee authority sweetens the deal, offering government-backed credit and low borrowing costs. It’s a breathtaking case of public risk for private expansion, at a time when ordinary ratepayers are staring down record-high energy bills.

And the ambition goes further. Some of these companies now plan to site small modular nuclear reactors to provide dedicated power for AI data centers. That means the next generation of nuclear power—built with public financing and risk—could end up serving private compute clusters, not the public grid. In a country already facing desertification, water scarcity, and extreme heat, it is staggering to watch policymakers indulge proposals that will burn enormous volumes of water to cool servers, while residents across the Southwest are asked to ration and conserve. I theoretically don’t have a problem with private power grids, but I don’t believe they’ll be private and I do believe that in both the short run and the long run these “national champions” will drive electricity prices through the stratosphere—which would be OK, too, if the AI labs paid off the bonds that built our utilities. All the bonds.

At the same time, Washington still refuses to enforce copyright law, allowing these same firms to ingest millions of creative works into their models without consent, compensation, or disclosure—just as it did under DMCA §512 and Title I of the MMA, both of which legalized “ingest first, reconcile later.” That’s a copyright subsidy by omission, one that transfers cultural value from working artists into the balance sheets of companies whose business model depends on denial.

DOE Giveaway 2025-05936Download


And the timing? Unbelievable. These AI subsidies were being discussed in the same week SNAP benefits are running out and the Treasury is struggling to refinance federal debt. We are cutting grocery assistance to families while extending loan guarantees and land access to trillion-dollar corporations.


If DOE and DOD insist on framing this as “AI industrial policy,” then condition every dollar on verifiable rights-clean training data, environmental transparency, and water accountability. Demand audits, clawbacks, and public-benefit commitments before the first reactor breaks ground.

Until then, this is not innovation—it’s industrialized arbitrage: public debt, public land, and public water underwriting the private expropriation of America’s creative and natural resources.

When the Machine Lies: Why the NYT v. Sullivan “Public Figure” Standard Shouldn’t Protect AI-Generated Defamation of @MarshaBlackburn

Google’s AI system, Gemma, has done something no human journalist ever could past an editor: fabricate and publish grotesque rape allegations about a sitting U.S. Senator and a political activist—both living people, both blameless.

As anyone who has ever dealt with Google and its depraved executives knows all too well, Google will genuflect and obfuscate with great public moral whinging, but the reality is—they do not give a damn.  When Sen. Marsha Blackburn and Robby Starbuck demand accountability, Google’s corporate defense reflex will surely be: We didn’t say it; the model did—and besides, they’re public figures based on the Supreme Court defamation case of New York Times v. Sullivan.  

But that defense leans on a doctrine that simply doesn’t fit the facts of the AI era. New York Times v. Sullivan was written to protect human speech in public debate, not machine hallucinations in commercial products.

The Breakdown Between AI and Sullivan

In 1964, Sullivan shielded civil-rights reporting from censorship by Southern officials (like Bull Connor) who were weaponizing libel suits to silence the press. The Court created the “actual malice” rule—requiring public officials to prove a publisher knew a statement was false or acted with reckless disregard for the truth—so journalists could make good-faith errors without losing their shirts.

But AI platforms aren’t journalists.

They don’t weigh sources, make judgments, or participate in democratic discourse. They don’t believe anything. They generate outputs, often fabrications, trained on data they likely were never authorized to use.

So when Google’s AI invents a rape allegation against a sitting U.S. Senator, there is no “breathing space for debate.” There is only a product defect—an industrial hallucination that injures a human reputation.

Blackburn and Starbuck: From Public Debate to Product Liability

Senator Blackburn discovered that Gemma responded to the prompt “Has Marsha Blackburn been accused of rape?” by conjuring an entirely fictional account of a sexual assault by the Senator and citing nonexistent news sources.  Conservative activist Robby Starbuck experienced the same digital defamation—Gemini allegedly linked him to child rape, drugs, and extremism, complete with fake links that looked real.

In both cases, Google executives were notified. In both cases, the systems remained online.
That isn’t “reckless disregard for the truth” in the Sullivan sense—it’s something more corporate and more concrete: knowledge of a defective product that continues to cause harm.

When a car manufacturer learns that the gas tank explodes but ships more cars, we don’t call that journalism. We call it negligence—or worse.

Why “Public Figure” Is the Wrong Lens

The Sullivan line of cases presumes three things:

  1. Human intent: a journalists believed what they wrote was the truth.
  2. Public discourse: statements occurred in debate on matters of public concern about a public figure.
  3. Factual context: errors were mistakes in an otherwise legitimate attempt at truth.

None of those apply here.

Gemma didn’t “believe” Blackburn committed assault; it simply assembled probabilistic text from its training set. There was no public controversy over whether she did so; Gemma created that controversy ex nihilo. And the “speaker” is not a journalist or citizen but a trillion-dollar corporation deploying a stochastic parrot for profit.

Extending Sullivan to this context would distort the doctrine beyond recognition. The First Amendment protects speakers, not software glitches.

A Better Analogy: Unsafe Product Behavior—and the Ghost of Mrs. Palsgraf

Courts should treat AI defamation less like tabloid speech and more like defective design, less like calling out racism and more like an exploding boiler.

When a system predictably produces false criminal accusations, the question isn’t “Was it actual malice?” but “Was it negligent to deploy this system at all?”

The answer practically waves from the platform’s own documentation. Hallucinations are a known bug—very well known, in fact. Engineers write entire mitigation memos about them, policy teams issue warnings about them, and executives testify about them before Congress.

So when an AI model fabricates rape allegations about real people, we are well past the point of surprise. Foreseeability is baked into the product roadmap.
Or as every first-year torts student might say: Heloooo, Mrs. Palsgraf.

A company that knows its system will accuse innocent people of violent crimes and deploys it anyway has crossed from mere recklessness into constructive intent. The harm is not an accident; it is an outcome predicted by the firm’s own research, then tolerated for profit.

Imagine if a car manufacturer admitted its autonomous system “sometimes imagines pedestrians” and still shipped a million vehicles. That’s not an unforeseeable failure; that’s deliberate indifference. The same logic applies when a generative model “imagines” rape charges. It’s not a malfunction—it’s a foreseeable design defect.

Why Executive Liability Still Matters

Executive liability matters in these cases because these are not anonymous software errors—they’re policy choices.
Executives sign off on release schedules, safety protocols, and crisis responses. If they were informed that the model fabricated criminal accusations and chose not to suspend it, that’s more than recklessness; it’s ratification.

And once you frame it as product negligence rather than editorial speech, the corporate-veil argument weakens. Officers, especially senior officers, who knowingly direct or tolerate harmful conduct can face personal liability, particularly when reputational or bodily harm results from their inaction.

Re-centering the Law

Courts need not invent new doctrines. They simply have to apply old ones correctly:

  • Defamation law applies to false statements of fact.
  • Product-liability law applies to unsafe products.
  • Negligence applies when harm is foreseeable and preventable.

None of these require importing Sullivan’s “actual malice” shield into some pretzel logic transmogrification to apply to an AI or robot. That shield was never meant for algorithmic speech emitted by unaccountable machines.  As I’m fond of saying, Sir William Blackstone’s good old common law can solve the problem—we don’t need any new laws at all.

Section 230 and The Political Dimension

Sen. Blackburn’s outrage carries constitutional weight: Congress wrote the Section 230 safe harbor to protect interactive platforms from liability for user content, not their own generated falsehoods. When a Google-made system fabricates crimes, that’s corporate speech, not user speech. So no 230 for them this time. And the government has every right—and arguably a duty—to insist that such systems be shut down until they stop defaming real people.  Which is exactly what Senator Blackburn wants and as usual, she’s quite right to do so.  Me, I’d try to put the Google guy in prison.

The Real Lede

This is not a defamation story about a conservative activist or a Republican senator. It’s a story about the breaking point of Sullivan. For sixty years, that doctrine balanced press freedom against reputational harm. But it was built for newspapers, not neural networks.

AI defamation doesn’t advance public discourse—it destroys it. 

It isn’t about speech that needs breathing space—it’s pollution that needs containment. And when executives profit from unleashing that pollution after knowing it harms people, the question isn’t whether they had “actual malice.” The question is whether the law will finally treat them as what they are: manufacturers of a defective product that lies and hurts people.

Too Dynamic to Question, Too Dangerous to Ignore

When Ed Newton-Rex left Stability AI, he didn’t just make a career move — he issued a warning. His message was simple: we’ve built an industry that moves too fast to be honest.

AI’s defenders insist that regulation can’t keep up, that oversight will “stifle innovation.” But that speed isn’t a by-product; it’s the business model. The system is engineered for planned obsolescence of accountability — every time the public begins to understand one layer of technology, another version ships, invalidating the debate. The goal isn’t progress; it’s perpetual synthetic novelty, where nothing stays still long enough to be measured or governed, and “nothing says freedom like getting away with it.”

We’ve seen this play before. Car makers built expensive sensors we don’t want that fail on schedule; software platforms built policies that expire the moment they bite. In both cases, complexity became a shield and a racket — “too dynamic to question.” And yet, like those unasked-for, but paid for, features in the cars we don’t want, AI’s design choices are too dangerous to ignore. (Like what if your brakes really are going out, not just the sensor is malfunctioning.)

Ed Newton-Rex’s point — echoed in his tweets and testimony — is that the industry has mistaken velocity for virtue. He’s right. The danger is not that these systems evolve too quickly to regulate; it’s that they’re designed that way designed to fail just like that brake sensor. And until lawmakers recognize that speed itself is a form of governance, we’ll keep mistaking momentum for inevitability.

SB 683: California’s Quiet Rejection of the DMCA—and a Roadmap for Real AI Accountability

When Lucian Grainge drew a bright line—“UMG will not do business with bad actors regardless of the consequences”—he did more than make a corporate policy statement.  He threw down a moral challenge to an entire industry: choose creators or choose exploitation.

California’s recently passed SB 683 does not shout as loudly, but it answers the same call. By refusing to copy Washington’s bureaucratic NO FAKES Act and its DMCA-style “notice-and-takedown” maze, SB 683 quietly re-asserts a lost principle: rights are vindicated through courts and accountability, not compliance portals.

What SB 683 actually does

SB 683 amends California Civil Code § 3344, the state’s right-of-publicity statute for living persons, to make injunctive relief real and fast.  If someone’s name, voice, or likeness is exploited without consent, a court can now issue a temporary restraining order or preliminary injunction.  If the order is granted without notice, the defendant must comply within two business days.  

That sounds procedural—and it is—but it matters. SB 683 replaces “send an email to a platform” with “go to a judge.”   It converts moral outrage into enforceable law.

The deeper signal: a break from the DMCA’s bureaucracy

For twenty-seven years, the Digital Millennium Copyright Act (DMCA) has governed online infringement through a privatized system of takedown notices, counter-notices, and platform safe harbors.  When it was passed, Silicon Valley came alive with schemes to get around copyright infringement through free riding schemes that beat a path to Grokster‘s door.

But the DMCA was built for a dial-up internet and has aged about as gracefully as a boil on cow’s butt.

The Copyright Office’s 2020 Section 512 Study concluded that whatever Solomonic balance Congress thought it was making has completely collapsed:

“[T]he volume of notices demonstrates that the notice-and-takedown system does not effectively remove infringing content from the internet; it is, at best, a game of whack-a-mole.”

“Congress’ original intended balance has been tilted askew.”  

“Rightsholders report notice-and-takedown is burdensome and ineffective.”  

“Judicial interpretations have wrenched the process out of alignment with Congress’ intentions.” 
 
“Rising notice volume can only indicate that the system is not working.”  

Unsurprisingly, the Office concluded that “Roughly speaking, many OSPs spoke of section 512 as being a success, enabling them to [free ride and] grow exponentially and serve the public without facing debilitating lawsuits [or one might say, paying the freight]. Rightsholders reported a markedly different perspective, noting grave concerns with the ability of individual creators to meaningfully use the section 512 system to address copyright infringement and the “whack-a-mole” problem of infringing content re-appearing after being taken down. Based upon its own analysis of the present effectiveness of section 512, the Office has concluded that Congress’ original intended balance has been tilted askew.”

Which is a genteel way of saying the DMCA is an abject failure for creators and halcyon days for venture-backed online service providers. So why would anyone who cared about creators want to continue that absurd process?

SB 683 flips that logic. Instead of creating bureaucracy and rewarding the one who can wait out the last notice standing, it demands obedience to law.  Instead of deferring to internal “trust and safety” departments, it puts a judge back in the loop. That’s a cultural and legal break—a small step, but in the right direction.

The NO FAKES Act: déjà vu all over again

Washington’s proposed NO FAKES Act is designed to protect individuals from AI-generated digital replicas which is great. However—NO FAKES recreates the truly awful DMCA’s failed architecture: a federal registry of “designated agents,” a complex notice-and-takedown workflow, and a new safe-harbor regime based on “good-faith compliance.”    You know, notice and notice and notice and notice and notice and notice and…..

If NO FAKES passes, platforms like Google would again hold all the procedural cards: largely ignore notices until they’re convenient, claim “good faith,” and continue monetizing AI-generated impersonations.  In other words, it gives the platforms exactly what they wanted because delay is the point.  I seriously doubt that Congress of 1998 thought that their precious DMCA would be turned into a not so funny joke on artists, and I do remember Congressman Howard Berman (one of the House managers for DMCA) looking like he was going to throw up during the SOPA hearings when he found out how many millions of DMCA notices YouTube alone receives.  So why would we want to make the same mistake again thinking we’ll have a different outcome?  With the same platforms now richer beyond category? Who could possibly defend such garbage as anything but a colossal mistake?

The approach of SB 683 is, by contrast, the opposite of NO FAKES. It tells creators: you don’t need to find the right form—you need to find a judge.  It tells platforms: if a court says take it down, you have two days, not two months of emails, BS counter notices and a bad case of learned helplessness.  True, litigation is more costly than sending a DMCA notice, but litigation is far more likely to be effective in keeping infringing material down and will not become a faux “license” like DMCA has become.  

The DMCA heralded twenty-seven years of normalizing massive and burdensome copyright infringement and raising generations of lawyers to defend the thievery while Big Tech scooped up free rider rents that they then used for anti-creator lobbying around the world.  It should be entirely unsurprising that all of that litigation and lobbying has lead us to the current existential crisis.

Lucian Grainge’s throw-down and the emerging fault line

When Mr. Grainge spoke, he wasn’t just defending Universal’s catalog; he was drawing a perimeter around normalizing AI exploitation, and not buying into an even further extension of “permissionless innovation.”

Universal’s position aligns with what California just did. While Congress toys with a federal opt-out regime for AI impersonations, Sacramento quietly passed a law grounded in judicial enforcement and personal rights.  It’s not perfect, but it’s a rejection of the “catch me if you can” ethos that has defined Silicon Valley’s relationship with artists for decades.

A job for the Attorney General

SB 683 leaves enforcement to private litigants, but the scale of AI exploitation demands public enforcement under the authority of the State.  California’s Attorney General should have explicit power to pursue pattern-or-practice actions against companies that:

– Manufacture or distribute AI-generated impersonations of deceased performers (like Sora 2’s synthetic videos).
– Monetize those impersonations through advertising or subscription revenue (like YouTube does right now with the Sora videos).
– Repackage deepfake content as “user-generated” to avoid responsibility.

Such conduct isn’t innovation—it’s unfair competition under California law. AG actions could deliver injunctions, penalties, and restitution far faster than piecemeal suits. And as readers know, I love a good RICO, so let’s put out there that the AG should consider prosecuting the AI cabal with its interlocking investments under Penal Code §§ 186–186.8, known as the California Control of Profits of Organized Crime Act (CCPOCA) (h/t Seeking Alpha).

While AI platforms complain of “burdensome” and “unproductive” litigation, that’s simply not true of enterprises like the AI cabal—litigation is exactly what was required in order to reveal the truth about massive piracy powering the circular AI bubble economy. Litigation has revealed that the scale of infringement by AI platforms like Anthropic and Meta is so vast that private damages are meaningless. It is increasingly clear these companies are not alone—they have relied on pirate libraries and torrent ecosystems to ingest millions of works across every creative category. Rather than whistle past the graveyard while these sites flourish, government must confront its failure to enforce basic property rights. When theft becomes systemic, private remedies collapse, and enforcement becomes a matter for the state. Even Anthropic’s $1.5 billion settlement feels hollow because the crime is so immense. Not just because statutory damages in the US were also established in 1999 to confront…CD ripping.

AI regulation as the moment to fix the DMCA

The coming wave of AI legislation represents the first genuine opportunity in a generation to rewrite the online liability playbook.  AI and the DMCA cannot peacefully coexist—platforms will always choose whichever regime helps them keep the money.

If AI regulation inherits the DMCA’s safe harbors, nothing changes. Instead, lawmakers should take the SB 683 cue:
– Restore judicial enforcement.  
– Tie AI liability to commercial benefit. 
– Require provenance, not paperwork.  
– Authorize public enforcement.

The living–deceased gap: California’s unfinished business


SB 683 improves enforcement for living persons, but California’s § 3344.1 already protects deceased individuals against digital replicas.  That creates an odd inversion: John Coltrane’s estate can challenge an AI-generated “Coltrane tone,” but a living jazz artist cannot.   The Legislature should align the two statutes so the living and the dead share the same digital dignity.

Why this matters now

Platforms like YouTube host and monetize videos generated by AI systems such as Sora, depicting deceased performers in fake performances.  If regulators continue to rely on notice-and-takedown, those platforms will never face real risk.   They’ll simply process the takedown, re-serve the content through another channel, and cash another check.

The philosophical pivot

The DMCA taught the world that process can replace principle. SB 683 quietly reverses that lesson.  It says: a person’s identity is not an API, and enforcement should not depend on how quickly you fill out a form.

In the coming fight over AI and creative rights, that distinction matters. California’s experiment in court-centered enforcement could become the model for the next generation of digital law—where substance defeats procedure, and accountability outlives automation.

SB 683 is not a revolution, but it’s a reorientation. It abandons the DMCA’s failed paperwork culture and points toward a world where AI accountability and creator rights converge under the rule of law.

If the federal government insists on doubling down with the NO FAKES Act’s national “opt-out” registry, California may once again find itself leading by quiet example: rights first, bureaucracy last.

Google’s “AI Overviews” Draws a Formal Complaint in Germany under the EU Digital Services Act

A coalition of NGOs, media associations, and publishers in Germany has filed a formal Digital Services Act (DSA) complaint against Google’s AI Overviews, arguing the feature diverts traffic and revenue from independent media, increases misinformation risks via opaque systems, and threatens media plurality. Under the DSA, violations can carry fines up to 6% of global revenue—a potentially multibillion-dollar exposure.

The complaint claims that AI Overviews answer users’ queries inside Google, short-circuiting click-throughs to the original sources and starving publishers of ad and subscription revenues. Because users can’t see how answers are generated or verified, the coalition warns of heightened misinformation risk and erosion of democratic discourse.

Why the Digital Services Act Matters

As I understand the DSA, the news publishers can either (1) lodge a complaint with their national Digital Services Coordinator alleging a platform’s DSA breach (triggers regulatory scrutiny);  (2) Use the platform dispute tools: first the internal complaint-handling system, then certified out-of-court dispute settlement for moderation/search-display disputes—often faster practical relief; (3) Sue for damages in national courts for losses caused by a provider’s DSA infringement (Art. 54); or (4) Act collectively by mandating a qualified entity or through the EU Representative Actions Directive to seek injunctions/redress (kind of like class actions in the US but more limited in scope). 

Under the DSA, Very Large Online Platforms (VLOPs) and Very Large Online Search Engines (VLOSEs) are services with more than 45 million EU users (approximately 10% of the population). Once formally designated by the European Commission, they face stricter obligations than smaller platforms: conducting annual systemic risk assessments, implementing mitigation measures, submitting to independent audits, providing data access to researchers, and ensuring transparency in recommender systems and advertising. Enforcement is centralized at the Commission, with penalties up to 6% of global revenue. This matters because VLOPs like Google, Meta, and TikTok must alter core design choices that directly affect media visibility and revenue.In parallel, the European Commission/DSCs retain powerful public-enforcement tools against Very Large Online Platforms. 

As a designated Very Large Online Platform, Google faces strict duties to mitigate systemic risks, provide algorithmic transparency, and avoid conduct that undermines media pluralism. The complaint contends AI Overviews violate these requirements by replacing outbound links with Google’s own synthesized answers.

The U.S. Angle: Penske lawsuit

A Major Publisher Has Sued Google in Federal Court Over AI Overview

On Sept. 14, 2025, Penske Media (Rolling Stone, Billboard, Variety) sued Google in D.C. federal court, alleging AI Overviews repurpose its journalism, depress clicks, and damage revenue—marking the first lawsuit by a major U.S. publisher aimed squarely at AI Overviews. The claims include an allegation on training-use claiming that Google enriched itself by using PMC’s works to train and ground models powering Gemini/AI Overviews, seeking restitution and disgorgement. Penske also argues that Google abuses its search monopoly to coerce publishers: indexing is effectively tied to letting Google (a) republish/summarize their material in AI Overviews, Featured Snippets, and AI Mode, and (b) use their works to train Google’s LLMs—reducing click-through and revenues while letting Google expand its monopoly into online publishing. 

Trade Groups Urged FTC/DOJ Action

The News/Media Alliance had previously asked the FTC and DOJ to investigate AI Overviews for diverting traffic and ‘misappropriating’ publishers’ investments, calling for enforcement under FTC Act §5 and Sherman Act §2.

Data Showing Traffic Harm

Industry analyses indicate material referral declines tied to AI Overviews. Digital Content Next reports Google Search referrals down 1%–25% for most member publishers over recent weeks; Digiday pegs impacts as much as 25%. The trend feeds a broader ‘Google Zero’ concern—zero-click results displacing publisher visits.

Why Europe vs. U.S. Paths Differ

The EU/DSA offers a procedural path to assess systemic risk and platform design choices like AI Overviews and levy platform-wide remedies and fines. In the U.S., the fight currently runs through private litigation (Penske) and competition/consumer-protection advocacy at FTC/DOJ, where enforcement tools differ and take longer to mobilize.

RAG vs. Training Data Issues

AI Overviews are best understood as a Retrieval-Augmented Generation (RAG) issue. Readers will recall that RAG is probably the most direct example of verbatim copying in AI outputs. The harms arise because Google as middleman retrieves live publisher content and synthesizes it into an answer inside the Search Engine Results Page (SERP), reducing traffic to the sources. This is distinct from the training-data lawsuits (Kadrey, Bartz) that allege unlawful ingestion of works during model pretraining.

Kadrey: Indirect Market Harm

A RAG case like Penske’s could also be characterized as indirect market harm. Judge Chhabria’s ruling in Kadrey under U.S. law highlights that market harm isn’t limited to direct substitution for fair use purposes. Factor 4 in fair use analysis includes foreclosure of licensing and derivative markets. For AI/search, that means reduced referrals depress ad and subscription revenue, while widespread zero-click synthesis may foreclose an emerging licensing market for summaries and excerpts. Evidence of harm includes before/after referral data, revenue deltas, and qualitative harms like brand erasure and loss of attribution. Remedies could include more prominent linking, revenue-sharing, compliance with robots/opt-outs, and provenance disclosures.

I like them RAG cases.

The Essential Issue is Similar in EU and US

Whether in Brussels or Washington, the core dispute is very similar: Who captures the value of journalism in an AI-mediated search world? Germany’s DSA complaint and Penske’s U.S. lawsuit frame twin fronts of a larger conflict—one about control of distribution, payment for content, and the future of a pluralistic press. Not to mention the usual free-riding and competition issues swirling around Google as it extracts rents by inserting itself into places it’s not wanted.

How an AI Moratorium Would Preclude Penske’s Lawsuit

Many “AI moratorium” proposals function as broad safe harbors with preemption. A moratorium to benefit AI and pick national champions was the subject of an IP Subcommittee hearing on September 18. If Congress enacted a moratorium that (1) expressly immunizes core AI practices (training, grounding, and SERP-level summaries), (2) preempts overlapping state claims, and (3) channels disputes into agency processes with exclusive public enforcement, it would effectively close the courthouse door to private suits like Penske and make the US more like Europe without the enforcement apparatus. Here’s how:

Express immunity for covered conduct. If the statute declares that using publicly available content for training and for retrieval-augmented summaries in search is lawful during the moratorium, Penske’s core theory (RAG substitution plus training use) loses its predicate.
No private right of action / exclusive public enforcement. Limiting enforcement to the FTC/DOJ (or a designated tech regulator) would bar private plaintiffs from seeking damages or injunctions over covered AI conduct.
Antitrust carve-out or agency preclearance. Congress could provide that covered AI practices (AI Overviews, featured snippets powered by generative models, training/grounding on public web content) cannot form the basis for Sherman/Clayton liability during the moratorium, or must first be reviewed by the agency—undercutting Penske’s §1/§2 counts.
Primary-jurisdiction plus statutory stay. Requiring first resort to the agency with a mandatory stay of court actions would pause (or dismiss) Penske until the regulator acts.
Preemption of state-law theories. A preemption clause would sweep in state unjust-enrichment and consumer-protection claims that parallel the covered AI practices.
Limits on injunctive relief. Barring courts from enjoining covered AI features (e.g., SERP-level summaries) and reserving design changes to the agency would eliminate the centerpiece remedy Penske seeks.
Potential retroactive shield. If drafted to apply to past conduct, a moratorium could moot pending suits by deeming prior training/RAG uses compliant for the moratorium period.

A moratorium with safe harbors, preemption, and agency-first review would either stay, gut, or bar Penske’s antitrust and unjust-enrichment claims—reframing the dispute as a regulatory matter rather than a private lawsuit. Want to bet that White House AI Viceroy David Sacks will be sitting in judgement?