NextFin News - US Attorney General Todd Blanche says the Justice Department will not police the artificial intelligence industry through aggressive enforcement, rejecting what he calls "regulation by prosecution" as the world's most valuable technology sector faces a fresh internal revolt over how fast its products should reach the public.
In an exclusive interview published September 15, 2026, Blanche criticized the prior administration's heavy-handed approach toward AI companies and executives, saying the department under his leadership will act only where there are clear violations of existing law. The remarks land as AI-linked equities wobble: Nvidia fell more than 2% in premarket trading on September 14 after a group of leading technology executives called for a slower pace of development on safety grounds, while Nasdaq 100 futures dropped 1.72% in early trading. The combination is the story - the industry is asking itself to slow down, and the government's top law enforcement officer is simultaneously signaling that criminal prosecutors will not fill the resulting safety gap.
The Doctrine: 'Regulation by Prosecution' Has a Paper Trail
Blanche's September 15 position did not appear out of nowhere. The phrase "regulation by prosecution" traces to a memorandum he issued as deputy attorney general on April 7, 2025, titled "Ending Regulation By Prosecution." In it, Blanche wrote of the digital assets sector:
The Department of Justice is not a digital assets regulator. However, the prior Administration used the Justice Department to pursue a reckless strategy of regulation by prosecution, which was ill conceived and poorly executed.
The mechanism is straightforward, and it matters for every regulated-adjacent industry. Regulation by prosecution occurs when the Justice Department uses criminal or civil enforcement actions to establish rules for an industry in the absence of legislation or agency rulemaking. Each case becomes a de facto rule, announced after the conduct has already occurred. The alternative is ex ante rules written by regulators or Congress, with enforcement limited to clear violations of those rules.
Under the 2025 memo, the department stopped targeting virtual currency exchanges, mixing and tumbling services, and offline wallets for the acts of their end users or unwitting regulatory violations. It disbanded the National Cryptocurrency Enforcement Team and narrowed its focus to prosecuting individuals who victimize investors or use digital assets to further crimes such as narcotics trafficking, terrorism financing, and fraud. The September 15 interview extends that same doctrine to artificial intelligence: prosecutors will not bring cases whose practical effect is to superimpose a regulatory framework on model developers, and will confine themselves to prosecutable harms - fraud, theft, exploitation, national-security breaches - committed with AI tools.
The timing of the message is deliberate. Blanche was sworn in as the 88th attorney general on August 10, 2026, after a narrow 50-49 Senate confirmation on August 8, ending months of service as acting attorney general that began in April 2026. His confirmation fight was bruising, with Democrats pressing him on his prior representation of President Trump and on whether the department would operate independently of the White House. A public commitment to enforce only clear laws is, among other things, a political shield: it is easier to defend "we follow the statute" than "we decided this novel technology should be treated as a crime."
The Market Is Pricing Safety Fears, Not Enforcement Risk
Here is the second-order point that most coverage of the selloff misses. Investors reacted to the September 14 safety warnings as a growth-and-spending story, not as a legal-risk story. The fear is that companies will voluntarily slow model releases and trim capital expenditure - a direct hit to revenue growth for chipmakers and cloud providers. At 04:46 a.m. ET that morning, Dow E-minis had shed 97 points, or 0.18%, S&P 500 E-minis were down 53.25 points, or 0.70%, and Nasdaq 100 E-minis had fallen 505.5 points, or 1.72%. Chipmakers took the hardest hit: Intel, AMD, and Marvell Technology were down roughly 6%, 5%, and 6%, respectively. Meta and Amazon each fell more than 1%.
But look at the other side of the tape. Shares of US software companies that have been dampened by fears that AI could disrupt their businesses actually climbed: ServiceNow, Adobe, and Workday rose about 3%, 2.5%, and 2.5%, respectively. The market is drawing a line between the companies that sell the frontier and the companies that sell into it. Slower frontier development protects the moats of enterprise software incumbents even as it threatens the growth narrative of the chipmakers.
Blanche's remarks add a second, independent signal to this mix. If the Justice Department will not regulate AI through the courts, then the regulatory risk premium embedded in AI valuations should, in theory, compress. That is bullish for the sector in the abstract. But it also means the safety question that sparked the selloff remains institutionally unresolved - and unresolved safety questions are what trigger the next wave of political pressure. A government that declines to regulate today may be forced to prosecute tomorrow, after a failure has already occurred. That is a worse outcome for everyone except trial lawyers.
The Governance Vacuum: If Not the Justice Department, Then Who?
With federal criminal enforcement pulling back, other actors are already moving into the space. The most consequential may be state attorneys general. Pennsylvania's State Board of Medicine filed suit on May 1, 2026, against Character Technologies, the operator of Character.ai, alleging that a user-created chatbot named "Emilie" unlawfully practiced medicine by posing as a licensed psychiatrist, providing a fabricated license number, and accumulating more than 45,000 user interactions by mid-April. The state sought a preliminary injunction - a regulatory outcome achieved through a century-old medical licensing statute rather than any AI-specific law.
At the federal level, consumer-protection enforcement has continued through the Federal Trade Commission, which secured a $193,000 settlement from the "robot lawyer" service DoNotPay over deceptive claims about its AI chatbot's ability to substitute for a human lawyer. And the courts have policed their own boundaries: in July 2025, a federal judge ordered two attorneys to each pay $3,000 after they submitted an AI-drafted court filing containing fabricated case citations.
Internationally, the US is exporting its hands-off philosophy. At the G20 Innovation Ministerial in Chapel Hill, North Carolina, on September 1, 2026, White House Office of Science and Technology Policy Director Michael Kratsios pressed other nations to avoid creating new rules for the technology.
Policymakers do not need to approach each innovation in isolation and should not treat every emerging technology as a first-of-its-kind policy problem.
The G20 leaders' summit is scheduled for Miami in December 2026, where AI governance will sit on the agenda alongside the US-China technology race.
Within the Justice Department itself, the antitrust line remains active even as the criminal line retreats. Acting Assistant Attorney General Omeed Assefi, speaking at New York University on May 7, 2026, said the division would continue to test merger-defense narratives that invoke AI as an industry-changing force.
We know you will be tempted to tell us that AI is replacing your industries. We get it. We hear that a lot. For us to take it seriously, we expect it to be backed up with actual evidence.
The division's November 24, 2025 proposed consent decree with rental-pricing software company RealPage - limiting its use of nonpublic competitor data - stands as a test case for algorithmic coordination. The message is nuanced: the department will not write new rules for AI through criminal indictments, but it will use existing antitrust tools where conduct fits established doctrine.
The Counter-Thesis: Restraint Is a Bet That the Harm Hasn't Happened Yet
The strongest argument against Blanche's approach is not that regulation is good in the abstract. It is that AI systems can produce diffuse, large-scale harm that no existing statute cleanly covers, and that a "clear violations only" policy means waiting for the harm to occur before the government can act.
The pressure for some form of guardrail is coming from inside the industry itself. On July 28, 2026, an open letter titled "Pacing the Frontier" was signed by more than 1,100 employees across OpenAI, Anthropic, Google DeepMind, and Meta - including senior scientists and co-founders - asking the US government to help build the technical and governance infrastructure for a coordinated slowdown if AI capability ever outruns human oversight. It is not a call for a pause today; it is a request for a brake pedal that no single company dares to press first. Demis Hassabis, speaking at the G20 gathering, went further and called for mandatory safety tests for AI systems.
The enforcement record shows the problem is real and already here. State officials are pursuing AI companies for chatbot harms under old statutes because no new ones exist. If the Justice Department declares itself not a regulator, Congress does not legislate, and agencies do not write rules, then the first enforcement action after a catastrophic failure will be both too late and politically explosive - precisely the "regulation by prosecution" dynamic Blanche says he wants to avoid, just delayed and intensified.
Blanche's answer, per the interview, is that the department will act where the law is clear. The counter-reply is that the law is precisely what is unclear for frontier AI systems, and that clarity is the very thing the private sector cannot create for itself.
Cyclical or Structural? This Is a Structural Shift With a Cyclical Overlay
Getting this call right determines the entire conclusion. The enforcement retreat is structural, not cyclical, for three reasons.
First, it is anchored in a written doctrine issued at the deputy attorney general level and now reaffirmed by the confirmed attorney general in his first major statement on the technology. Second, it is paired with institutional dismantling - the disbanding of the National Cryptocurrency Enforcement Team shows the administration is willing to remove enforcement infrastructure, not merely slow its use. Third, it aligns with the administration's broader deregulatory program: the revocation of the prior administration's AI executive orders, the "Winning the AI Race" action plan released in July 2025, and the G20 hands-off push.
A structural enforcement retreat will not self-correct. It reverses only with a change of administration or a catalytic event severe enough to force legislation. Investors should treat lower federal enforcement risk as a durable feature of this administration, not a mood that will pass.
But there is a cyclical overlay, and it is what is moving prices today. The market's reaction to safety warnings is a sentiment swing that can reverse quickly. A single strong earnings report, a breakthrough model release, or a shift in the tone of executive commentary could restore risk appetite within days - and the selloff would be revealed as noise layered on top of a structural trend. The two forces point in different directions and operate on different clocks: structurally, enforcement risk for AI companies is lower for the remainder of this administration; cyclically, AI valuations remain vulnerable to sentiment shifts around safety and capital expenditure.
What Comes Next: Beneficiaries, the Exposed, and the Signal That Would Prove This Wrong
Who benefits is clear. AI model developers and chipmakers gain a lower enforcement risk premium, and the United States preserves a home-field advantage in the race with China for AI supremacy. Enterprise software companies benefit if frontier development slows and their domain-specific products gain relative appeal. Who is exposed is equally clear: investors betting on AI safety as a regulatory catalyst; companies that relied on clear federal guardrails to justify compliance spending; and state attorneys general, who may find themselves overextended as the default AI police force in a federal vacuum.
Three things to watch. First, the G20 leaders' summit in Miami in December 2026 - whether the US hands-off line holds when other nations press for binding rules. Second, Congress - if it fails to pass AI legislation before the 2026 midterm elections, the vacuum persists into the next cycle. Third, the next major AI incident. A fatality, a large-scale fraud, or a national-security breach involving AI would test the "clear violations only" doctrine within hours, not months.
Scenarios, not a single line. The base case is no major AI catastrophe before the midterms, federal enforcement staying restrained, rapid industry deployment continuing, and valuations remaining volatile but trending higher on earnings. The upside case is a breakthrough model or strong earnings cycle that restores momentum, vindicating the hands-off posture as US companies extend their lead over China. The downside case is a high-profile AI failure that triggers public outrage, prompting Congress or state attorneys general to impose a fragmented patchwork of rules that turns the enforcement vacuum into a political liability.
The falsifying signal is specific and observable: if the Justice Department brings a novel criminal enforcement action against an AI company for conduct not clearly prohibited by an existing statute within the next six months, the "no regulation by prosecution" doctrine is weaker than stated. Watch the department's charging documents, not its press releases.
Blanche is betting that American AI dominance is worth more than the safety that regulation might buy - and that the market, not the courtroom, will decide where the line gets drawn. It is a coherent position. It is also a position that can be falsified by a single catastrophe, and that is the bet the entire sector is now riding on.
Explore more exclusive insights at nextfin.ai.
