What happened on 29 September

On 29 September 2026 the President of the United States signed Executive Order 14434, "Inaugurating the Era of Super Intelligence". It was published in the Federal Register on 2 October 2026 at 91 FR 63129 (FR Doc. 2026-20321). The order runs to two pages and four sections. Its central instruction is a change of vocabulary: federal executive departments and agencies are to write "Super Intelligence" and "SI" wherever they have so far written "Artificial Intelligence" and "AI".

The same afternoon, after a White House meeting with technology executives, the President and senior figures from six companies signed a short voluntary document, reported as the "White House Accord on Super Intelligence". Five days later, on 4 October, the President announced in a social media post a new federal body he called the "Super Intelligence Force", to be led by the Director of National Intelligence, Jay Clayton.

The three were presented as a package, but their legal standing differs. The order is a published legal instrument. The accord's text is known only from press reports, and the task force so far exists only in a social media post and press accounts. This article keeps them apart.

What the order says, section by section

Section 1 (Purpose) gives the policy rationale. It says today's frontier systems "do much more than imitate or automate discrete aspects of human intelligence", and that "Super Intelligence" better captures what they can do. It then declares that, "to the maximum extent permitted by law", the executive branch is to use the new terms in place of the old ones and "will not acknowledge the usage" of "Artificial Intelligence" and "AI" "in any applicable setting".

Section 2 (Implementation) is the operative instruction. Under section 2(a), agencies are to use "Super Intelligence" and "SI" in official correspondence, public communications, websites, reports, policy documents "and other non-statutory documents within the executive branch". Section 2(b) sets an important limit: nothing in the section requires changes to previously issued regulations, Presidential actions, contracts, grants or other historical documents.

Section 3 (Definition) does two things. Section 3(a) provides that, for the purposes of the order and except where the law provides otherwise, "Super Intelligence" and "SI" mean the technologies and systems covered by "artificial intelligence" as defined in 15 U.S.C. § 9401(3). That definition was introduced by the National Artificial Intelligence Initiative Act of 2020. It describes a machine-based system that can, for a given set of human-defined objectives, make predictions, recommendations or decisions that influence real or virtual environments. The new label therefore covers exactly the same technologies as the old one.

Section 3(b) gives the Assistant to the President for Science and Technology (APST) 60 days to send the President proposed legislative language for a federal definition of "Super Intelligence" and "SI" that reflects the capabilities described in section 1. The proposal must include:

  • an assessment of whether the new definition should modify, expand on or supersede the existing statutory definition of "artificial intelligence";
  • any conforming amendments to existing statutory references; and
  • recommendations for further Presidential or executive action.

Sixty days from 29 September is 28 November 2026. The APST is currently Michael Kratsios, who also directs the Office of Science and Technology Policy (OSTP).

Section 4 (General provisions) contains the standard clauses: the order is implemented consistently with applicable law and subject to available funding, and it "is not intended to, and does not, create any right or benefit, substantive or procedural, enforceable at law or in equity by any party".

The White House fact sheet of the same day summarises the order in the same terms and describes a set of legislative recommendations unveiled in March 2026 as the "National Super Intelligence Policy Framework". It does not mention the accord or any task force.

What the order does not do

Because the order is so short, it is easy to read more into it than is there. On its own text:

  • It does not change the law. An executive order directs the executive branch; it cannot amend a statute. The definition in 15 U.S.C. § 9401(3), and every statute that uses "artificial intelligence", stay as they are. The order itself acknowledges this by asking for legislative language, which only Congress can enact.
  • It does not change scope. Section 3(a) ties "SI" to the existing statutory definition of AI. No technology enters or leaves any regulatory category because of this order.
  • It imposes nothing on private parties. It does not tell companies, contractors or foreign governments to do anything, and section 4(c) rules out enforceable rights for anyone.
  • It does not rewrite existing documents. Under section 2(b), previously issued regulations, Presidential actions, contracts and grants are untouched.
  • It is silent on state law, export controls and safety. There is no preemption clause, no reference to the Export Administration Regulations, and no testing, licensing or reporting duty.

What it does do is administrative. Federal solicitations, guidance and websites can be expected to switch to "SI"; according to a Wiley Rein alert, NIST has said it is updating its communications as directed. The substantive change, if any, would come later, if Congress acts on the 28 November proposal.

One phrase in section 1 needs a closer look: the statement that the executive branch "will not acknowledge the usage" of "Artificial Intelligence" and "AI" "in any applicable setting". Section 2 confines the operative duty to the executive branch's own documents, and the whole policy is subject to the "maximum extent permitted by law" qualifier. The most natural reading is that the phrase governs how agencies write, not how they treat documents written by others. Nothing in the order makes a bid or a public comment non-compliant because it says "AI".

The accord and the task force: what has been reported

The accord. The following account comes from press reports by Euronews, Forbes Australia, Al Jazeera and Nextgov. The document signed on 29 September is reported as the "White House Accord on Super Intelligence", subtitled "Joint Commitment on Frontier Responsibilities"; one report treats the subtitle as a companion text. The President signed it together with:

  • Dario Amodei (Anthropic);
  • Sundar Pichai (Google);
  • Mark Zuckerberg (Meta);
  • Greg Brockman (OpenAI);
  • Jensen Huang (Nvidia); and
  • Elon Musk (xAI).

As reported, the text is just over 300 words long. It describes four layers of oversight for companies that train and deploy frontier models:

  1. internal controls to monitor model capabilities and alignment;
  2. an internal team empowered to check those controls;
  3. an independent external auditor or evaluator; and
  4. an independent committee of the board of directors to oversee the reports.

It adds that "over time, it may make sense to codify these steps into laws or regulations".

The reports agree on what the accord leaves out: no penalties, no enforcement role for the government, no deadline, and no duty to name the auditor or publish audit results. Asked whether it binds anyone, the President said he thought it was "morally binding". Microsoft and Amazon, both among the seven companies that gave voluntary commitments to the previous administration in July 2023, are not among the reported signatories. According to Euronews, the accord was posted on the President's social media account. We could not find it on whitehouse.gov, and the fact sheet does not mention it, so its reported wording cannot yet be checked against an official source.

The "Super Intelligence Force". On 4 October, according to the Associated Press and TechCrunch, the President said in a social media post that Jay Clayton, the Director of National Intelligence, would lead a task force called the "Super Intelligence Force". Its stated purpose is to coordinate federal efforts "to ensure that America continues to lead the World in Super Intelligence". AP names three other members: FTC Chairman Andrew Ferguson, the Pentagon's chief technology officer Emil Michael, and the Director of the Office of Personnel Management, Scott Kupor. It reports that the group answers to the President and his chief of staff.

TechCrunch, citing the Wall Street Journal, reports that the body has 120 days to report on the risks and opportunities of the technology, and that its charter refers both to responding to "SI-enabled threats" and to avoiding "overregulation and regulatory capture". We could not find any charter or establishing instrument, and as of 9 October the Federal Register contained no presidential document about it. Its legal basis, membership and timetable are therefore provisional.

Where the order sits in federal AI policy

EO 14434 is the latest in a line of presidential AI orders that changed course in January 2025.

  • EO 14110 (30 October 2023) was the previous administration's order on "Safe, Secure, and Trustworthy" AI. It required developers of the largest models to report to the government. EO 14148 revoked it on 20 January 2025.
  • EO 14179 (23 January 2025), "Removing Barriers to American Leadership in Artificial Intelligence", set the present policy direction. OMB implemented it in April 2025 with memoranda on agencies' use and procurement of AI (M-25-21 and M-25-22).
  • EOs 14318, 14319 and 14320 (23 July 2025) accompanied America's AI Action Plan. They dealt with data-centre permitting, with the ideological neutrality of models procured by federal agencies, and with an "American AI Exports Program" for full-stack US technology packages.
  • EO 14365 (11 December 2025), "Ensuring a National Policy Framework for Artificial Intelligence", is the preemption order. It:
    • told the Attorney General to set up an AI Litigation Task Force to challenge state AI laws;
    • told the Commerce Department to identify "onerous" state laws;
    • directed the Commerce Department to make states with such "onerous" AI laws ineligible, as far as federal law allows, for certain remaining federal broadband (BEAD) funds; and
    • called for a legislative recommendation for a uniform federal framework preempting conflicting state laws, leaving child safety, data-centre infrastructure and state procurement outside it.
  • EO 14409 (2 June 2026), "Promoting Advanced Artificial Intelligence Innovation and Security", focused on cyber defence. It directed agencies to design a voluntary framework under which developers could give the government access to "covered frontier models" before release. It says expressly that nothing in the relevant section authorises a mandatory licensing, preclearance or permitting requirement for new models.

Against that background, EO 14434 is legally the lightest order in the series. Its significance lies in two places.

First, with the accord, it continues the approach of EO 14409: governance through voluntary arrangements with industry rather than binding rules. Speaking the same day, Vice President Vance rejected an FDA- or FAA-style regulator for frontier models, according to Nextgov. He pointed instead to the existing powers of the FTC and the Justice Department over consumer harms.

Second, the 28 November proposal reopens the federal statutory definition of the technology. The order does not say whether that proposal will be linked to the preemption legislation envisaged by section 8 of EO 14365. Some commentators have raised the possibility, but for now it is speculation.

The states moved in the same fortnight. According to a release by the Washington Attorney General dated 24 September, a bipartisan coalition of 26 attorneys general, including those of the District of Columbia and American Samoa, wrote to congressional leaders asking for, among other things, federal oversight of safety testing and "a prohibition on preemption of state laws". On 30 September, California's Governor signed 13 further bills, 12 of them on AI, covering, among other things, lawyers' use of AI and AI-only disciplinary and dismissal decisions at work.

Under the US Constitution, preemption of state law comes from federal statutes and valid federal regulations. An executive order alone cannot displace a state statute. For now, state AI laws remain in force unless Congress legislates over them or a court strikes them down.

What it means for companies outside the United States

Federal procurement. Some non-US suppliers sell software, cloud or analytics services to US federal agencies, either directly or through US prime contractors. They should expect "SI" in solicitations, questionnaires and agency guidance. Section 3(a) ties "SI" to the statutory definition of AI, so the scope of procurement requirements does not change: a requirement that applied to an "AI system" applies equally to an "SI system". Existing contracts are unaffected (section 2(b)). The practical step is to make sure internal compliance mapping recognises both terms, so that no requirement is missed because it carries a different label. It is not yet known whether OMB will reissue its April 2025 memoranda in the new vocabulary.

Export controls. The order does not touch the Export Administration Regulations or any sanctions programme. US controls on advanced computing items are set by technical performance parameters, not by the word "AI". Nothing changes for non-US companies that buy US chips or models; the relevant export-promotion policy is still EO 14320's American AI Exports Program. Freshfields notes that a future statutory definition could shape export controls and other federal programmes. That is a possibility to watch, not a present effect.

Contracts. The order does not change private contracts. Where a US-law contract incorporates the statutory definition by reference, the text is still today's 15 U.S.C. § 9401(3). If Congress later amends it, contracts referring to the statute "as amended" could change in scope; those that reproduce the definition word for word would not. That is worth checking in long-term agreements with US counterparties.

The accord. Non-US companies are not parties and owe nothing under it. Its four layers are, however, close to what customers already ask of AI suppliers, and companies that supply the signatories may see the same expectations passed down through contract terms.

The EU, Korean and Turkish frameworks

The renaming has no legal effect outside the United States. Each of the frameworks that matter most to readers of this site defines the technology in its own terms.

European Union. The AI Act (Regulation (EU) 2024/1689) defines an "AI system" in Article 3(1) and applies to providers placing systems on the EU market wherever they are established, so a US company's systems remain "AI systems" in the EU whatever Washington calls them. The more instructive contrast is one of method: the EU attaches binding, risk-based obligations and fines to the technology, while the US federal executive relies on voluntary commitments and, in EO 14409, expressly rules out licensing.

Korea. Korea's AI Basic Act (Framework Act on the Development of Artificial Intelligence and the Establishment of a Foundation for Trust) has been in force since 22 January 2026. It has its own statutory definition of artificial intelligence, places transparency and safety duties on certain AI business operators, and requires foreign operators that meet thresholds set by Presidential Decree to designate a domestic representative in Korea. None of this is affected by the US renaming.

Türkiye. Türkiye has no general AI statute in force. Its main current policy instrument is the AI Action Plan (2026–2030), put into effect by Presidential Circular 2026/9 in the Official Gazette of 18 August 2026. The Plan is addressed to public bodies and does not itself impose obligations on private parties. The US order changes nothing in Turkish law. The terminology of Turkish legislation and policy documents, including the Action Plan and obligations under Law No. 6698 on the Protection of Personal Data, is unaffected. For Turkish exporters to the US federal market, the only effect is the procurement-vocabulary point above.

What to watch next

  • 28 November 2026: the APST's proposed legislative language for a federal definition of "Super Intelligence" is due to the President. The order does not require it to be published.
  • Congress: any statutory change needs legislation.
  • The task force: whether a formal instrument is published, and when the reported 120-day report arrives. Counting from 4 October, that would be early February 2027.
  • The accord: whether an official text is published, whether other companies sign, and whether any signatory names its external auditor or board committee.
  • Preemption: whether the definitional proposal is tied to federal preemption legislation, and how that interacts with the new California laws and the attorneys general's demand that state law be preserved.

Frequently asked questions

Has "artificial intelligence" been removed from US law?

No. The order binds only the executive branch's own non-statutory documents. Statutes, existing regulations, contracts and grants keep the term "artificial intelligence". A statutory change would need an Act of Congress, and the order only asks for a proposal.

Does a Turkish or Korean company need to rename its products or rewrite its documents?

No. Non-US law is unaffected, and US federal law has not changed its definition. Companies that sell to US federal agencies should make sure their compliance teams treat "SI" and "AI" as the same thing in solicitations and questionnaires, and should check how their US-law contracts refer to statutory definitions.

Is the accord legally binding?

On the reported text, no. It is voluntary, has no penalties and gives the government no enforcement role. The President described it as "morally binding". Its official text has not been located on the White House website.

Burhan Doğuş Ayparlar's View

This section sets out my personal assessment as the founder of this site and an attorney at law (Türkiye).

It would be easy to treat this order as a matter of words and move on, and that would be half right. EO 14434 is drafted with care: section 3(a) anchors the new term to the existing statutory definition, section 2(b) leaves existing contracts, grants and regulations untouched, and section 4(c) rules out new rights. For now it is a style instruction to the federal executive, and a company outside the United States need only make sure its people recognise the new label.

The part that deserves attention is section 3(b). In technology law, definitions are not cosmetic: they decide which systems fall inside an obligation and which fall outside it. The current US definition in 15 U.S.C. § 9401(3) and the EU definition in Article 3(1) of the AI Act are close cousins. Both describe machine-based systems that produce outputs such as predictions, recommendations or decisions for given objectives and influence an environment. That shared structure is what lets a company build one compliance file and use it across jurisdictions. If the 28 November proposal leads to a statutory definition built around capability rather than function, the overlap could shrink. Organisations that sell in both markets would then carry a translation cost. Whether that happens depends on the proposal's content and on Congress, and both are unknown today.

On the accord, I would separate the substance from the setting. The four layers it describes are internal controls, an internal team that checks them, an external auditor and a board committee. Taken together they are a familiar governance structure, and well-run companies in any sector would recognise it. What determines the value of such commitments is verifiability. As reported, the text names no auditors, sets no dates and requires no disclosure, so outsiders will not be able to tell whether it is being followed. Voluntary commitments can still shape practice, especially when they flow into supply contracts.

For Turkish and Korean organisations, the measured response is a modest one. Map "SI" to "AI" in any US federal procurement work. Check how US-law contracts refer to statutory definitions. Watch what is proposed on 28 November and what, if anything, Congress does with it. The obligations that bind them today come from the EU AI Act, Korea's AI Basic Act and Turkish data-protection and sectoral rules, and none of those changed on 29 September.

This article is for information only and does not constitute legal advice. The description of Executive Order 14434 is based on its text as published in the Federal Register on 2 October 2026 (91 FR 63129) and the White House fact sheet of 29 September 2026. The descriptions of the "White House Accord on Super Intelligence" and the "Super Intelligence Force" are based on press reports published between 29 September and 5 October 2026; no official text of either was located as of 9 October 2026, and details may change. Statements about US, EU, Korean and Turkish law are general in nature; specific cases require individual assessment. The analysis and assessments are the author's own.