The memorandum is three pages. It is addressed to all Department of Justice employees, signed by the Attorney General, dated 9 January 2026, and titled — with the unembellished functionality of an internal directive — Artificial Intelligence Litigation Task Force. In the space of those three pages it quotes Executive Order 14365 five separate times, establishes a new structure within the Department whose sole responsibility is to challenge state AI laws, names the Attorney General or her designee as Chair and the Associate Attorney General as Vice Chair, lists the offices that will compose the Task Force, and identifies the White House officials with whom the Task Force shall “consult from time to time.” It does almost nothing else. The brevity is the document’s most informative feature. An entire arm of federal law enforcement has just been created to litigate against the states, and the instrument doing the creating reads like a personnel notice.
↑ N° 11 · Continues themes from N° 11. The executive order this memorandum implements — Ensuring a National Policy Framework for Artificial Intelligence — concedes in its own Section 8 that the framework it announces requires a Congress that has twice declined to enact one. This piece reads the implementing memorandum the order produced inside the Justice Department.What the memo actually does
The operative text is short enough to quote in full without it dominating the article. It is also short enough that the choice of brevity becomes the editorial signal — the memorandum is a performance of fidelity to the executive order it implements.
The body of the memorandum opens by quoting the President directly. “United States leadership in Artificial Intelligence (AI) will promote United States national and economic security and dominance across many domains.” That sentence is Section 1 of Executive Order 14365, signed on 11 December 2025. The next quotation — “United States AI companies must be free to innovate without cumbersome regulation,” followed by the patchwork-of-fifty argument — is also from Section 1. The third — that the policy of the United States is “to sustain and enhance the United States’ global AI dominance through a minimally burdensome national policy framework for AI” — is from Section 2. The fourth — “establish[ing] an AI Litigation Task Force (Task Force) whose sole responsibility shall be to challenge State AI laws inconsistent with the policy” — is from Section 3. The fifth, naming the consultative White House officials, is also from Section 3.
Five quotations. Five footnotes. The footnotes read, in their entirety, Executive Order 14365 § 1, Id., Id. § 2, Id. § 3, Id. The Attorney General has written the memorandum as a near-transcription of the order. The grounds on which the Task Force will challenge state laws — “that such laws unconstitutionally regulate interstate commerce, are preempted by existing Federal regulations, or are otherwise unlawful” — are themselves verbatim from the order. The only operative language the Attorney General contributes that is not directly traceable to the executive order is the composition list and the chair designations.
The choice not to elaborate is the choice that does the most work in the document. The Attorney General could have written, in the same memorandum, an interpretive paragraph on what counts as “inconsistent with the policy,” or a procedural section on how the Task Force will identify candidate state laws, or a delegation paragraph clarifying which Deputy or Associate Deputy will run the Task Force in practice. None of that is in the memorandum. The Department’s leadership has chosen instead to let the executive order’s text do the work and to let the institutional fact of the Task Force’s existence carry the policy signal.
A task force of one job
The phrase that does the most work in the executive order — repeated by the memorandum — is “sole responsibility.” Federal task forces almost never have one. The choice of an exclusive mandate is itself a policy instrument.
The Department of Justice operates dozens of task forces. The Election Threats Task Force, established in 2021, handles threats to election workers but also coordinates with U.S. Attorneys’ Offices on a portfolio of related election-integrity matters. The Foreign Corrupt Practices Act Unit’s enforcement work fits inside a broader Fraud Section that prosecutes other kinds of fraud. The Cybersecurity Unit in the National Security Division supports prosecutions across multiple statutes. The institutional default is overlapping responsibilities — a task force coordinates resources that already exist elsewhere in the Department.
The AI Litigation Task Force, by contrast, has a single mandate, written into its founding document, that admits of no other work. Its sole responsibility shall be to challenge state AI laws. There are no other items in the brief. There is no second-order responsibility to advise on federal AI policy, no responsibility to coordinate with other federal agencies on AI matters, no responsibility to monitor compliance, no responsibility to defend federal AI initiatives against legal challenge. The Task Force exists to do one thing.
The composition list reinforces the point. The Task Force draws from the Office of the Deputy Attorney General (the operational leadership of the Department), the Office of the Associate Attorney General (which oversees the civil litigating divisions), the Office of the Solicitor General (which controls appellate strategy and Supreme Court litigation), and the Civil Division (which actually files and litigates civil cases on the United States’ behalf). These are the four offices that, together, can take a constitutional challenge to a state law from initial filing through Supreme Court argument without involving anyone else. The Task Force is, in organisational terms, a fully self-contained federal litigation pipeline pointed at the states.
The Attorney General reserved the right to designate “any other components or representatives that I may from time to time designate to assist in the Task Force’s work.” That is the only flexibility in the memorandum. It contemplates expansion, not redirection.
The grounds, and what they don't include
Three theories. The memorandum lists them. The Department’s first actual filing relied on only one — and a different one from any the memorandum lists. The gap between the announced grounds and the deployed grounds is the doctrinal news.
The grounds the Task Force may invoke are “that such laws unconstitutionally regulate interstate commerce, are preempted by existing Federal regulations, or are otherwise unlawful.” The first is the Dormant Commerce Clause — the implied limit on state authority to regulate interstate commerce that runs through cases such as Pike v. Bruce Church and, more recently, National Pork Producers v. Ross. The second is statutory or regulatory preemption — the doctrine that a state law conflicting with a federal statute or regulation must yield, under the Supremacy Clause and the framework the Supreme Court has built in Geier v. American Honda Motor and its successors. The third — “otherwise unlawful” — is a residual category that could mean almost anything.
The first actual filing under the Task Force’s umbrella was the Department’s intervention against the Colorado AI Act in xAI Corp. v. Weiser. The intervention complaint, filed on 24 April 2026 by Acting Attorney General Todd Blanche under his Section 902 certification authority, rested on neither of the first two grounds the memorandum lists. It rested on the Equal Protection Clause of the Fourteenth Amendment — the theory that a state law requiring AI developers and deployers to mitigate algorithmic discrimination, with an explicit carve-out for measures designed to advance “diversity” or “redress historical discrimination,” itself compels discrimination on the basis of race and other protected characteristics.
SB24-205 is unconstitutional in other ways too. To comply with SB24-205, developers and deployers must make different editorial decisions regarding training data, responses to prompts, model constraints, and more — all to generate Colorado’s preferred expressive outputs.
— xAI Corp. v. Weiser, Complaint in Intervention
The doctrinal turn is worth pausing on. The memorandum lists three plausible grounds for challenging state AI laws. The Department’s first concrete deployment used a fourth. The Equal Protection theory inverts the conventional understanding of anti-discrimination law: a state law explicitly aimed at preventing discriminatory outputs is recharacterised as a law compelling discrimination, on the basis of its carve-outs for protected characteristics. The theory’s success or failure in the District of Colorado, and on the appellate review that is almost certain to follow, will define the doctrinal terrain on which all subsequent Task Force actions are litigated.
↑ N° 04 · Continues themes from N° 04. The Department’s intervention in xAI v. Weiser is the Task Force’s first concrete action. The litigation has since produced a joint motion suspending enforcement while the Colorado legislature considers a replacement bill.The memorandum’s three listed grounds remain available. The Dormant Commerce Clause theory — that state AI regulation impermissibly reaches conduct outside the state’s borders — was raised by xAI in the underlying complaint and is plausible against many state AI laws that purport to regulate developers regardless of where they are located. The preemption theory is constrained by the fact that there is no federal AI statute to preempt anything; the Task Force will need to identify federal regulations under existing statutes — the Federal Trade Commission Act, the Civil Rights Acts, sectoral statutes in finance, housing, employment — and argue that state AI laws conflict with the federal regulatory schemes those statutes authorise. The “otherwise unlawful” category remains undefined.
The consultative list, and what it signals
The memorandum names four White House officials with whom the Task Force shall consult. The list is short and unusual. It explains the policy architecture the Task Force is built to serve.
The four consultative officials are the Special Advisor for AI and Crypto; the Assistant to the President for Science and Technology; the Assistant to the President for Economic Policy; and the Assistant to the President and Counsel to the President.
The Special Advisor for AI and Crypto is David Sacks, the venture capitalist and longtime Trump ally appointed to the new role in December 2024. Sacks does not have a confirmed position in the federal hierarchy — the Special Advisor role is a White House staff position that did not exist before this administration created it — and he is not a lawyer. His inclusion at the top of the consultative list signals that the policy direction of the Task Force flows through him to the Department, rather than through the more traditional channels of OMB review or interagency coordination.
The Assistant to the President for Science and Technology is the head of the Office of Science and Technology Policy, a position with statutory grounding under the National Science and Technology Policy Act of 1976. The Assistant to the President for Economic Policy heads the National Economic Council, established by Executive Order 12835 of 1993. The Assistant to the President and Counsel to the President is the senior in-house legal advisor to the presidency and the conduit for presidential legal positions across the executive branch.
The absent names are equally informative. The memorandum does not list the Attorney General’s traditional litigation counterparts — the General Counsels of the Federal Trade Commission, the Federal Communications Commission, the Department of Commerce, or any of the sectoral regulators with existing AI-relevant authority. It does not list the Solicitor of Labor, the General Counsel of the Equal Employment Opportunity Commission, or the General Counsels of the financial regulators. The consultative list is exclusively White House. The Task Force will receive its policy direction from the political offices of the presidency, not from the federal regulatory agencies whose statutes the preemption theory will most heavily depend upon.
What the memorandum creates, and what it leaves to the courts
The memorandum is a complete institutional act and an incomplete legal one. It establishes a Task Force. It does not establish a legal theory that has yet survived appellate review.
What the memorandum has created, on the day it was signed, is irreversible without further executive action. The AI Litigation Task Force exists. Its leadership is named. Its composition is set. Its consultative architecture is established. The Department of Justice has, in three pages, redirected institutional resources toward a litigation programme against the states that did not exist before 9 January 2026.
What it has not created is durable legal doctrine. The Task Force’s first deployment, in xAI v. Weiser, is currently in suspended enforcement while the Colorado legislature drafts a replacement statute. The Equal Protection theory the Department advanced has not been tested at the district court level on a contested record, let alone at the Tenth Circuit or above. The Dormant Commerce Clause theory has not been advanced. The preemption theory cannot be advanced until federal regulations are in place that the state laws could be argued to conflict with — and the Federal Communications Commission proceeding and the Federal Trade Commission policy statement directed by Sections 6 and 7 of Executive Order 14365 have not yet produced operative documents.
The memorandum’s most significant disclosure is its date. The Task Force was established on 9 January 2026, twenty-nine days after the executive order was signed and twenty-six days after it was published in the Federal Register. The Department moved quickly. The other federal agencies directed by the same order have not — the Commerce Department’s March 11 evaluation deadline has passed without a public document, the Federal Communications Commission proceeding has not been initiated, the Federal Trade Commission policy statement has not been issued. The Justice Department’s institutional response, signed by the Attorney General within a month of the executive order, is the only piece of the executive order’s architecture that has so far been fully implemented.
What the reader is left with is the institutional commitment behind a three-page document. The Attorney General signed it. The Task Force exists. The first case is filed. The legal theory has not yet won. The framework the executive order announced has not yet been written. The implementing memorandum has done the most concrete work the executive order has produced in five months — and it has done it by quoting the order’s text rather than by adding to it. The shape of federal AI policy in the United States, for the moment, is the shape of a litigation programme run out of the Department of Justice on five citations to a single executive order.