AI is now “SI” in Washington. What actually changed?

For most of the world, the technology behind chatbots and coding assistants is still called artificial intelligence. Inside the US federal government, it now has a different official name. This guide sums up what happened, what the new rule actually requires, and what it does not do.

How it happened

In mid-September 2026, President Trump asked his followers on social media to vote on new names for AI, with options such as superior, extreme and supreme intelligence. On 22 September he told the UN General Assembly that the United States would call the technology super intelligence, or SI.

A week later, on 29 September, he hosted around twenty technology leaders at a White House lunch and signed an executive order titled Inaugurating the Era of Super Intelligence. Trump has argued that the word “artificial” makes the technology sound fake, and described it as bigger than the internet.

What the order requires

• Departments and agencies of the executive branch use “Super Intelligence” and “SI” in official communications, public documents and policy documents, and stop using “artificial intelligence” and “AI”.
• The rule applies as far as the law allows. Agencies do not have to rewrite regulations, contracts or records that already exist.
• The President’s science and technology adviser has 60 days to propose a federal definition of super intelligence.
• Until that definition exists, SI simply covers whatever the existing federal legal definition of AI already covers.

What it does not change

The order binds the federal executive branch. It does not apply to Congress, the courts, companies, researchers or other countries, so nobody else is required to switch terms.

It is also not a scientific claim. In research, superintelligence usually means a hypothetical system that clearly outperforms humans across nearly all intellectual work. Today’s systems are not that, and several computer scientists and ethicists have said the new label overstates what current AI can do. Some have suggested the term should be kept for a much more advanced level of AI.

Adoption outside government is mixed so far. Some figures close to the administration have switched, and Elon Musk corrected himself from “AI” to “SI” on camera at the White House, but most companies and researchers still say AI.

The accord signed the same day

Before the order was signed, leaders from Google, Anthropic, Meta, OpenAI, SpaceX and NVIDIA signed a separate one-page document, the White House Accord on Superintelligence. According to reporting based on CNBC, it asks companies to set up internal safety oversight, have it assessed by external auditors, and send reports to an independent committee.

A side effect: the rush for .si domains

Slovenia’s country code, .si, suddenly reads as “super intelligence”. The Slovenian registry counted 46,066 new .si registrations in September and more than 13,000 within a single day at the start of October. Reporting so far suggests that most buyers are domain investors hoping to resell, rather than companies building products under .si.

Does this change the forecast?

No. Renaming the technology does not make superintelligence arrive sooner.

The markets on SafeASI ask whether superintelligence in the research sense will exist by a certain year. Their resolution criteria are about what systems can do, not about what a government calls them. A federal document that says “SI” today is talking about ordinary AI, so it cannot resolve any of these markets early.

If you want to see how the term itself relates to AGI and ASI, read AGI, ASI and SI: what the words actually mean.