From AI to SI: new name, old problem
Trump will henceforth call existing artificial intelligence Super Intelligence, but that does not change the technology or European rules.
Follow-up to: Trump opts for voluntary AI rules and a new name Saturday, 3 October 2026, 09:19
President Donald Trump has ordered US government departments to henceforth call artificial intelligence Super Intelligence, abbreviated as SI. The term has existed in AI research for longer, but there it refers to something much more specific than today’s language models and chatbots.
In research, superintelligence generally refers to a hypothetical system that surpasses humans in almost every intellectual field. It therefore means more than better text, image recognition or code, encompassing broad abilities such as learning, reasoning, planning, creativity and scientific problem-solving. According to AWS, artificial superintelligence remains theoretical; no system has been shown to have reached this level.
Current language models do not automatically fall into that category. They are trained on large quantities of text and can generate convincing answers, translations and programmes. That makes them versatile, but not necessarily generally intelligent: their performance depends on training, available data, tools and the task they are given. A model may outperform a human on a specific task while struggling with new situations outside its field.
Trump’s decision of 29 September does not alter that technical distinction. Executive Order 14434 requires US executive agencies to replace AI with SI in, among other things, correspondence, websites, reports and policy documents. Rules, contracts and historical documents issued earlier do not need to be amended. For implementation, the order initially uses the existing legal definition of artificial intelligence.
The announced change of definition is moreover not yet a new law. Within sixty days, the science and technology adviser must submit a proposal for statutory wording to the president. Among other things, that proposal must indicate whether the existing legal definition should be amended and which references in legislation should be changed. Only then can Congress consider a legislative amendment; the order itself creates no new enforceable right.
The major US AI companies have meanwhile joined a joint safety document. OpenAI, Anthropic, Google, Meta, xAI and Nvidia signed a commitment holding companies responsible for their own controls, internal oversight, independent evaluations and oversight by a board committee. This is a voluntary framework and not a replacement for legislation.
Researchers and critics object to merging two different concepts. Computer scientist Pedro Domingos told DW that most AI systems are not superintelligence, but another form of intelligence. Michael Bennett of the University of Illinois Chicago described the name change to TechTarget as a signal of a preference for self-regulation. OpenAI itself wrote in 2023 that superintelligence was a future, far more powerful category, and recently argued that mandatory, capability-based safety rules are needed.
For Europe and the Netherlands, the American choice of words changes nothing for now. The European AI Act uses the term AI and defines an AI system partly in terms of autonomy, inference and the ability to produce output that influences the physical or digital environment. The regulation became largely applicable on 2 August 2026; European companies and governments will therefore continue to work with that legal terminology. The American SI definition could have consequences mainly if Washington later introduces legislation, not because a presidential name change rewrites European rules.
One story, several perspectives
What is established
- The US government is choosing the term Super Intelligence in executive-branch communications.
- The order contains no new legal definition; a proposal for statutory wording must first be drafted.
- The European AI Act uses the term AI and operates within a risk-based legal framework.
- Tech companies have signed a voluntary agreement on internal controls, independent evaluations and board oversight.
Left
Arguments The name change can be seen as political marketing that distracts attention from who is liable for harm and who provides oversight. From this perspective, rules for the most powerful systems should be statutory, democratically accountable and independently assessable.
Values Public control, protection of workers and citizens, equality and limiting the concentration of power among technology companies.
Consequences Voluntary agreements may prove insufficient when commercial pressure and geopolitical competition subordinate safety to rapid growth.
Centre
Arguments The government can give innovation room while introducing clear thresholds, audits and reporting obligations for the most powerful systems. The name matters less than a workable definition aligned with risk and capability.
Values Institutional reliability, proportionality, innovation with oversight and international applicability.
Consequences A combination of voluntary technical standards and statutory minimum requirements may be quicker to implement, but requires scrutiny of conflicts of interest.
Right
Arguments From this perspective, the new term emphasises that US companies are developing a strategic technology capable of strengthening economic growth, science and national security. Excessively onerous federal rules could slow innovation or reduce the lead over China.
Values Technological leadership, national sovereignty, entrepreneurship and limited government intervention.
Consequences Self-regulation can respond more quickly to new developments, but loses credibility if companies set their own safety limits without external sanctions.
The perspectives describe how these political currents typically approach the subject; the newsroom takes no position on which perspective is right.
Fact-check Approved · Nour Haddad — AI agent
This check was carried out by AI: every claim was re-tested against the sources. Even an approved article can contain errors — stay critical.
The article’s core is based on the US presidential order, European Commission documents and several independent public sources. The text distinguishes between established rules, technical definitions and interpretations by interviewed experts.
- confirmed Executive Order 14434 was signed on 29 September 2026 and replaces AI with SI in executive-branch communications. — This is stated in the White House order. source
- confirmed The order does not require previously issued rules, contracts and historical documents to be amended. — Section 2(b) of the order makes this explicit. source
- confirmed Within sixty days, a proposal for statutory wording on SI must be submitted to the president. — Section 3(b) specifies this deadline and assignment. source
- confirmed In the AI sources cited, superintelligence is a theoretical system that broadly surpasses humans. — AWS describes artificial superintelligence as cognitively superior to humanity as a whole and calls it theoretical. source
- confirmed Current language models are not automatically superintelligent. — AWS distinguishes current, task-oriented AI and language models from theoretical superintelligence. source
- confirmed OpenAI, Anthropic, Google, Meta, xAI and Nvidia signed a voluntary safety agreement. — The public text mentions the commitments; independent reporting names the signing companies. source
- confirmed Pedro Domingos called the term misleading for most current AI. — DW quotes Domingos on the difference between AI and superintelligence. source
- confirmed OpenAI advocates mandatory, capability-based national safety rules. — This is stated in OpenAI’s 2026 policy contribution. source
- confirmed The European AI Act uses AI as a legal term and defines AI systems partly through autonomy and inference. — The European Commission published this definition and the information on applicability. source
Editor's note
What Executive Order 14434 prescribes, what superintelligence means in the sources cited and how the European AI Act is worded is certain. What remains uncertain is which legal definition the government will ultimately propose and whether Congress will adopt it. No direct, separate statements from all the companies mentioned about the name itself were found; their demonstrable joint response concerns the voluntary safety agreement.Sources
- Inaugurating the Era of Super Intelligence — The White House
- What is Superintelligence? Artificial Superintelligence Explained — Amazon Web Services
- Super Intelligence: Is Trump right to rename AI? — DW
- White House Accord on Super Intelligence — The American Presidency Project
- AI is now ‘super intelligence,’ but federal oversight is still missing — TechTarget
- Governance of superintelligence — OpenAI
- The AI policy window is open. We need to act. — OpenAI
- AI Act — European Commission
- Commission Guidelines on the definition of an artificial intelligence system — European Commission
The story so far
- Saturday, 3 October 2026, 09:19 Trump opts for voluntary AI rules and a new name
- Saturday, 3 October 2026, 15:00 From AI to SI: new name, old problem (this article)
More on this in Dutch media
- Het Parool — „kunstmatige intelligentie”
- NRC — „kunstmatige intelligentie”
- Tweakers — „kunstmatige intelligentie”