THEBLACKBOOK AUDIT
The Tech Right · The oldest tech-policy move, run again

The Section 230 playbook: asking to be untethered from the law itself.

Thirty years ago we decided the internet was “wholly different” and gave platforms broad immunity, preempting state law and cutting them loose from the common law built up over generations. The law professor Zephyr Teachout argues the AI industry is now running the same move — and the 2025 record largely backs her up.

This page tests Teachout's claim against what the industry is actually lobbying for right now, and grades it. The documented asks — federal preemption of state AI laws, a single national framework, a failed attempt to freeze state regulation for a decade — are real and on the record. Whether that adds up to a Section 230-style bid for “immunity from the law” is her sharp reading, and we grade it as strong analysis rather than settled fact. And because this is about AI rules, we disclose up top that the model writing it is made by a company with a stake in the fight — and that, unusually, that company's own position cuts against the loudest version of the ask.

The graded claims ↓Our conflict of interest
§0 · Conflict of Interest Disclosure

This page about AI regulation was drafted with a model made by an AI company — and we're precise about where that company stands

Black Book Audit uses Claude, built by Anthropic. This investigation is about what the AI industry is asking regulators for, so our maker is an interested party and we grade its interests harder, per The AI-Safety Money. But honesty cuts a specific way here: on the exact asks this page examines, Anthropic is not the aggressor. Its CEO publicly opposed the proposed decade-long freeze on state AI laws, and Anthropic endorsed California's SB 53 transparency law while other industry groups lobbied against it. So we will not lump Anthropic in with the preemption push to score an easy point — and we will not use Anthropic's comparatively better stance to launder the industry as a whole or to soften Teachout's critique. The company still wants a federal framework and still benefits from the broader “safety” positioning we scrutinize elsewhere. Name the split; grade each part on its own evidence.

§1 · Summary Brief

What this page argues

In 1996, Section 230 of the Communications Decency Act gave online platforms broad immunity from liability for what their users posted, and courts read it to preempt most state laws that would hold platforms responsible — treating “the internet” as a special category outside the common law. Zephyr Teachout's argument is that the AI industry, with different motives, is now asking for the same structural gift: to be untethered from state regulators, from existing federal law, and from tort — a “wholly new regime” instead of accountability under the law as it already exists.

The concrete asks are documented. OpenAI asked Congress in 2025 for federal preemption to escape a wave of state AI bills. Congress twice tried to attach a sweeping freeze on state AI regulation to must-pass legislation, and the Senate stripped a ten-year version 99–1. And in December 2025 the White House issued an executive order — driven by AI czar David Sacks — directing the government to identify and challenge state AI laws and to build “a minimally burdensome national standard, not 50 discordant State ones,” a move 36 bipartisan state attorneys general opposed and legal experts say likely exceeds executive power. We grade those facts as facts, and Teachout's reading of them — that preemption functions as a back-door immunity, stripping out the state tort and consumer-protection enforcement where accountability actually lives — as strong, probably-true analysis. The honest complication, disclosed above, is that the industry is not monolithic: our own maker opposed the freeze and backed a transparency law. The dominant legal ask of 2025, though, was exactly the escape-from-law move Teachout names.

What we are NOT claiming
We are not claiming that every AI company seeks total immunity — our own maker's record cuts the other way — or that a federal framework is inherently illegitimate; the industry's complaint about a fragmented 50-state patchwork is a real cost, not a fiction. We are not endorsing Teachout's specific policy prescription as proven correct; we present it as her argued alternative. What we document is that the industry's dominant 2025 legal ask — preemption of state law, a single national regime, and a failed moratorium — structurally echoes the Section 230 “wholly different” move, and that a bid for preemption is, in effect, a bid to shed the liabilities that state law imposes.
Recommended reading

Books that go deeper on this story. Links are Amazon affiliate searches — buying through them supports the work at no cost to you.

▶ Dossier

The same investigation, restaged one beat at a time. Step through it here, or present it fullscreen.

The Tech Right

The Section 230 playbook.

30 years ago we called the internet 'wholly different' and gave platforms immunity, preempting state law. Zephyr Teachout says the AI industry is running the same move — and the 2025 record backs her up.

1 / 9▶ Present fullscreen
§2 · Graded Claims

The record, claim by claim

The precedent is real: Section 230 gave platforms broad immunity and preempted state law, treating the internet as a special category.

FACT

Teachout's historical premise is accurate. Section 230 of the 1996 Communications Decency Act generally shields providers of interactive computer services from liability for content posted by others, and its subsection (e) has been read by courts to preempt most state laws that would impose such liability — creating a uniform federal standard and, in effect, carving the internet out of the ordinary common-law and state-law regime that governs other industries. Legal scholars across the spectrum note the preemptive scope has been applied expansively, well beyond what some argue the original text required. Whatever its benefits, Section 230 is the clearest modern example of an industry being placed in its own special legal category — which is exactly the precedent Teachout says is being reached for again.

Today's ask #1: federal preemption of state AI laws — requested by OpenAI, and pursued by a December 2025 White House executive order.

FACT

The central move Teachout names is documented. In 2025 OpenAI asked Congress for federal preemption to relieve it from the hundreds of AI bills introduced in the states. On December 11, 2025, President Trump issued an executive order advancing federal preemption of state AI laws — framed as building 'a minimally burdensome national standard, not 50 discordant State ones' — directing AI czar David Sacks and the Commerce Department to identify state AI laws that conflict with the policy and to challenge them, and to draft a unifying federal framework. A bipartisan coalition of 36 state attorneys general has opposed a federal ban on state AI laws, and legal analysts widely note that the executive branch likely cannot preempt state regulation this way without Congress passing a statute. The push to strip states out of the picture is not a prediction; it is the current policy.

Today's ask #2: the moratorium — Congress twice tried to freeze all state AI regulation, and the Senate stripped a ten-year version 99–1.

FACT

The most naked version of the 'wholly different regime' move was the proposed moratorium. In 2025, congressional Republicans twice attempted to attach a sweeping preemption of state AI laws — at one point a ten-year freeze — to must-pass legislation, which would have barred states from regulating AI at all while no federal framework existed to replace them. The Senate stripped the ten-year moratorium from its reconciliation bill in a 99–1 vote in July 2025. The attempt failed, but it is the clearest evidence of the industry-aligned goal Teachout describes: not a particular rule, but exemption from the states' power to make rules at all.

Teachout's reading: preemption is a back-door immunity — it strips out the state tort and consumer-protection enforcement where accountability actually lives.

PROBABLY TRUE

This is the interpretive heart of Teachout's argument, and we grade it as strong analysis rather than a documented demand. Her point: most real accountability for corporate harm happens through state law — tort suits, state consumer-protection statutes, and enforcement by state attorneys general acting for their citizens (parens patriae). Preempting those, or freezing states out entirely, does not merely 'harmonize' rules; it removes the venues where injured people and state enforcers can actually reach a company — functionally the same shield Section 230 gave platforms. Whether the industry frames its goal as 'immunity' or as 'a single national standard,' the effect Teachout identifies is real: a preemption regime with a weak or captured federal backstop leaves little law with teeth. We grade this PROBABLY TRUE — a sound reading of what preemption does — while noting that 'they want total immunity from tort' is her characterization of the effect, not a slogan the companies themselves use.

The honest complication: the industry is not monolithic — and our own maker is the counterexample.

FACT

Fairness and our own conflict require this block. The preemption-and-moratorium push has been driven hardest by venture and platform interests — Andreessen Horowitz, OpenAI's Washington operation, and the administration's David Sacks. But the AI industry does not speak with one voice on it. Anthropic — the maker of the model that wrote this page — publicly opposed the ten-year moratorium, with CEO Dario Amodei calling it 'too blunt an instrument' that would leave 'no ability for states to act, and no national policy as a backstop,' and Anthropic endorsed California's SB 53 transparency law while groups like the Consumer Technology Association lobbied against it. We report that straight, without inflating it: Anthropic still seeks a federal framework and still benefits from the safety-first positioning we scrutinize elsewhere, and endorsing disclosure rules is a smaller thing than accepting full tort exposure. But Teachout's 'the AI companies' is not uniform, and the honest version of her claim targets the dominant preemption push, not every firm.

Teachout's alternative, stated fairly: keep AI inside existing law, and build any federal regulator within an existing agency to resist capture.

PROBABLY TRUE

We present her prescription as the argued position it is, and grade its internal logic rather than stamping it as correct policy. Teachout argues that if AI's threats to health and wellbeing are taken seriously, the answer is more law, not less: preserve state law and tort, strengthen consumer-protection statutes with real private rights of action, use existing antitrust tools against corporate collusion, and rely on decentralized enforcement — including state attorneys general acting parens patriae — so that capture at the federal level does not become wholesale capture. Where a federal regulator is warranted, she argues for building it inside an existing agency (on the model of the Nuclear Regulatory Commission and the National Nuclear Security Administration) rather than standing up a new, single-purpose body that is easier to capture, both by industry and by what she calls the 'post-law ideology' of the effective-altruism world. Her core structural claim — that decentralization is a hedge against capture, and that preemption plus a single new regulator concentrates the very risk it claims to manage — is a sound governance argument, which is why we grade the logic PROBABLY TRUE even as the policy choices remain debatable.

  • Zephyr Teachout — the proposed 'precautionary regime': state law + tort + strengthened consumer protection + antitrust + decentralized (parens patriae) enforcement + a regulator built within an existing agency (NRC/NNSA model)
§3 · Record vs Narrative

Where the evidence is strong, and where it stops

  • The asks are documented. OpenAI's preemption request, the twice-attempted moratorium, and the December 2025 executive order are on the record — this is current policy, not speculation.
  • “Immunity” is the effect, described. The companies say “national standard,” not “immunity”; Teachout's contribution is showing that preemption of state law produces the immunity result. We grade that as analysis, and a strong one.
  • The industry is split. The preemption drive is led by specific players; our own maker opposed the moratorium and backed a transparency law. “The AI companies” is not uniform, and we say so.
  • The patchwork cost is real. We don't pretend 50 conflicting state regimes are costless. The question is who resolves that — a democratic federal statute, or preemption that mostly deletes accountability.
§4 · Why It Matters

The most powerful ask isn't for a rule — it's to be outside the rules

Teachout's insight is that the fight people are watching — what the AI rules should say — is downstream of a quieter fight over whether ordinary law applies at all. That's the connective tissue to The Wrong AI Debate, where the loud conversation about speculative risk crowds out the present harms, and to The AI-Safety Money, where “safety” framing doubles as an incumbency moat. A single federal regulator, especially a new one, is far easier to capture than fifty state legislatures, hundreds of tort juries, and dozens of state attorneys general — which is precisely why a decentralized system is a hedge, and why a preemption-plus-immunity regime concentrates risk under a banner of managing it. Section 230 is the cautionary tale: a “wholly different” carve-out, sold as nurturing a young industry, that hardened into a liability shield the largest companies on earth now guard fiercely. The question this time is whether we run the same experiment before we know the harms — which belongs in The Tech Right and the wider Fake Opposition story of rules written to look like guardrails and function like moats.

§5 · Questions

Questions worth taking seriously

Isn't a single national AI standard better than 50 conflicting state laws?

The patchwork problem is real — we don't pretend otherwise. But there are two very different ways to solve it. One is a federal statute, debated and passed by Congress, that sets a floor and preserves people's ability to sue for harm. The other is preemption — by executive order or a moratorium — that deletes state authority without putting a real, enforceable federal regime in its place, leaving injured people and state enforcers with nowhere to go. Teachout's point isn't that national coordination is bad; it's that “national standard” is being used as the label for the second thing, which is mostly the removal of accountability.

Can a Claude-written page really be trusted on what AI companies want from regulators?

Read it adversarially — that's the right instinct. Our answer is in the §0 disclosure: we name Anthropic as an interested party and grade its interests harder. Here that meant reporting a fact that helps Anthropic (it opposed the moratorium and backed a transparency law) without letting that exonerate the industry or soften Teachout's critique, and reporting the facts that don't flatter any AI company (the preemption push is real and dangerous). If anything, the tell to watch for is a Claude page going easy on the safety-regulation framing — so we flagged that too.

§6 · Standing Invitation

If you are named on this page

If you are named on this page, or are a party materially affected by the claims made here, and you wish to respond, correct the record, or add context, use the Contact page. Responses are published verbatim alongside the original claim, with the sender identified and the date of receipt. The channel stays open for the life of the page.

This site aggregates and grades a record that other outlets and primary sources have already put on the record. Every FACT-graded claim above is sourced to court filings, government reports, sworn whistleblower disclosures, published investigative journalism, or named-source statements. The citations are the accountability mechanism; this section is how you get on the record too.

§7 · Sources

The record

▦ Ledger gaps

Help us fill these lines.

This entry is graded on what’s on the public record. These are the blanks we know about. If you can source one, you’re rebuilding the ledger with us.

  • OpenIf the December 2025 executive order is struck down for exceeding executive power, does the industry pivot back to a congressional moratorium — and what would a genuine federal floor (that preserves private rights of action) look like versus preemption that mainly deletes liability?Help fill this →
  • OpenWhich specific state tort and consumer-protection actions against AI harms would a preemption regime foreclose — and who loses standing to sue?Help fill this →

Notify me when a gap is filled

We'll email you when we fill one of the gaps above.

By signing up you agree to receive emails from The Black Book Audit. Unsubscribe anytime.