By Chris Castle

The AI safe harbor slavered onto President Trump’s “big beautiful bill” is layered with intended consequences. Not the least of these is the affect on TikTok.

One of the more debased aspects of TikTok (and that’s a long list) is their promotion through their AI driven algorithms of clearly risky behavior to their pre-teen audience. Don’t forget: TikTok’s algorithm is not just any algorithm. The Chinese government claims it as a state secret. And when the CCP claims a state secret they ain’t playing. So keep that in mind.

One of these risky algorithms that was particularly depraved was called the “Blackout Challenge.” The TikTok “blackout challenge” has been linked to the deaths of at least 20 children over an 18-month period. One of the dead children was Nylah Anderson. Nylah’s mom sued TikTok for her daughter because that’s what moms do. If you’ve ever had someone you love hang themselves, you will no doubt agree that you live with that memory every day of your life. This unspeakable tragedy will haunt Nylah’s mother forever.

Even lowlifes like TikTok should have settled this case and it should never have gotten in front of a judge. But no–TikTok tried to get out of it because Section 230. Yes, that’s right–they killed a child and tried to get out of the responsibility. The District Court ruled that the loathsome Section 230 applied and Nylah’s mom could not pursue her claims. She appealed.

The Third Circuit Court of Appeals reversed and remanded, concluding that “Section 230 immunizes only information ‘provided by another’” and that “here, because the information that forms the basis of Anderson’s lawsuit—i.e., TikTok’s recommendations via its FYP algorithm—is TikTok’s own expressive activity, § 230 does not bar Anderson’s claims.”

So…a new federal proposal threatens to slam the door on these legal efforts: the 10-year artificial intelligence (AI) safe harbor recently introduced in the House Energy and Commerce Committee. If enacted, this safe harbor would preempt state regulation of AI systems—including the very algorithms and recommendation engines that Nylah’s mom and other families are trying to challenge. 

Section 43201(c) of the “Big Beautiful Bill” includes pork, Silicon Valley style, entitled the “Artificial Intelligence and Information Technology Modernization Initiative: Moratorium,” which states:

no State or political subdivision thereof may enforce any law or regulation regulating artificial intelligence models, artificial intelligence systems, or automated decision systems during the 10-year period beginning on the date of the enactment of this Act.

The “Initiative” also appropriates “$500,000,000, to remain available until September 30, 2035, to modernize and secure Federal information technology systems through the deployment of commercial artificial intelligence, the deployment of automation technologies, and the replacement of antiquated business systems….” So not only did Big Tech write themselves a safe harbor for their crimes, they also are taking $500,000,000 of corporate welfare to underwrite it courtesy of the very taxpayers they are screwing over. Step aside Sophocles, when it comes to tragic flaws, Oedipus Rex got nothing on these characters.

Platforms like TikTok, YouTube, and Instagram use AI-based recommendation engines to personalize and optimize content delivery. These systems decide what users see based on a combination of behavioral data, engagement metrics, and predictive algorithms. While effective for keeping users engaged, these AI systems have been implicated in promoting harmful content—ranging from pro-suicide material to dangerous ‘challenges’ that have directly resulted in injury or death.

Families across the country have sued these companies, alleging that the AI-driven algorithms knowingly promoted hazardous content to vulnerable users. In many cases, the claims are based on state consumer protection laws, negligence, or wrongful death statutes. Plaintiffs argue that the companies failed in their duty to design safe systems or to warn users about foreseeable dangers. These cases are not attacks on free speech or user-generated content; they focus specifically on the design and operation of proprietary AI systems. 

If you don’t think that these platforms are depraved enough to actually raise safe harbor defenses, just remember what they did to Nylah’s mom–raised the exceptionally depraved Section 230 as a defense to their responsibility in the death of a child.

The AI safe harbor would prohibit states from enacting or enforcing any law that regulates AI systems or automated decision-making technologies for the next 10 years. This sweeping language could easily be interpreted to cover civil liability statutes that hold platforms accountable for the harms their AI systems cause. This is actually even worse than the vile Section 230–the safe harbor would be expressly targeting actual state laws. Maybe after all the appeals, say 20 years from now, we’ll find out that the AI safe harbor is unconstitutional commandeering, but do we really want to wait to find out?

Because these wrongful death lawsuits rely on arguments that an AI algorithm caused harm—either through its design or its predictive content delivery—the companies could argue that the moratorium shields them from liability. They might claim that the state tort claims are an attempt to “regulate” AI in violation of the federal preemption clause. If courts agree, these lawsuits could be dismissed before ever reaching a jury.

This would create a stunning form of corporate immunity even beyond the many current safe harbors for Big Tech: tech companies would be free to deploy powerful, profit-driven AI systems with no accountability in state courts, even when those systems lead directly to preventable deaths. 

The safe harbor would be especially devastating for families who have already suffered tragic losses and are seeking justice. These families rely on state wrongful death laws to hold powerful platforms accountable. Removing that path to accountability would not only deny them closure, but also prevent public scrutiny of the algorithms at the center of these tragedies.

States have long held the authority to define standards of care and impose civil liability for harms caused by negligence or defective products. The moratorium undermines this traditional role by barring states from addressing the specific risks posed by AI systems, even in the context of established tort principles. It would represent one of the broadest federal preemptions of state law in modern history—in the absence of federal regulation of AI platforms.

• In Pennsylvania, the parents of a teenager who committed suicide alleged that Instagram’s algorithmic feed trapped their child in a cycle of depressive content.
• Multiple lawsuits filed under consumer protection and negligence statutes in states like New Jersey, Florida, and Texas seek to hold platforms accountable for designing algorithms that systematically prioritize engagement over safety.
• TikTok faced multiple class action multidistrict litigation claims it illegally harvested user information from its in-app browser.

All of such suits could be in jeopardy if courts interpret the AI moratorium as barring state laws that impose liability on algorithm-driven systems and you can bet that Big Tech platforms will litigate the bejeezus out of the issue. Even if the moratorium was not intended to block wrongful death and other state law claims, its language may be broad enough to do so in practice—especially when leveraged by well-funded corporate legal teams.

Even supporters of federal AI regulation should be alarmed by the breadth of this safe harbor. It is not a thoughtful national framework based on a full record, but a shoot-from-the-hip blanket prohibition on consumer protection and civil justice. By freezing all state-level responses to AI harms, the AI safe harbor is intent on consolidating power in the hands of federal bureaucrats and corporate lobbyists, leaving ordinary Americans with fewer options for recourse, not to mention a clear violation of state police powers and the 10th Amendment.

To add insult to injury, the use of reconciliation to pass this policy—without full hearings, bipartisan debate, or robust public input—only underscores the cynical nature of the strategy. It has nothing to do with the budget aside from the fact that Big Tech is snarfing down $500 million of taxpayer money for no good reason just so they can argue their land grab is “germane” to shoehorn it into reconciliation under the Byrd Rule. It’s a maneuver designed to avoid scrutiny and silence dissent, not to foster a responsible or democratic conversation about how AI should be governed.

At its core, the AI safe harbor is not about fostering innovation—it is about shielding tech platforms from accountability just like the DMCA, Section 230 and Title I of the Music Modernization Act. By preempting state regulation, it could block families from using long-standing wrongful death statutes to seek justice for the loss of their children and laws protecting Americans from other harms. It undermines the sovereignty of states, the dignity of grieving families, and the public’s ability to scrutinize the AI systems that increasingly shape our lives. 

Congress must reject this overreach, and the American public must remain vigilant in demanding transparency, accountability, and justice. The Initiative must go.

[A version of this post first appeared on MusicTechPolicy]