Nicole Jaffrey
When Donald Trump gathered artificial intelligence leaders at the White House, he wanted more than a discussion about technology’s future. He wanted a new name. For a president who understands the power of branding, “super intelligence” offers a stronger promise than “artificial intelligence.” But does that promise describe a scientific achievement, or anticipate one?
The September 29 meeting brought together executives including Mark Zuckerberg, Jensen Huang, Sundar Pichai, Dario Amodei and Greg Brockman. Trump announced a voluntary safety agreement covering internal controls and independent assessments, while reaffirming support for expanding data centres. The political message was clear: accelerate innovation while reassuring the public that its risks are being monitored.
That same day, he signed an executive order directing executive agencies to replace “Artificial Intelligence” and “AI” with “Super Intelligence” and “SI” in official communications and documents, where legally permitted. Crucially, the order retained the existing statutory definition of artificial intelligence pending further action. The terminology changed immediately; the technologies covered by it did not.
This exposes a gap between political language and scientific terminology. Artificial intelligence describes a broad field containing systems with widely varying capabilities. Superintelligence, in research discussions, refers to systems substantially smarter than humans. OpenAI’s research has treated that possibility as a challenge requiring new methods to keep more capable systems aligned with human intentions. Renaming existing technology does not demonstrate that this threshold has been reached.
Yet names can have economic consequences. “Artificial” describes how the technology originates; “super” suggests exceptional performance. That language may strengthen investment narratives and support infrastructure spending. It may also inflate expectations, encouraging users to trust systems beyond what independently verified results justify.
The safety agreement raises a more consequential question: who oversees the companies developing these systems? WIRED’s examination describes voluntary commitments involving internal monitoring, external assessments and board oversight. These measures may help, but they do not themselves constitute a comprehensive regulatory framework establishing responsibilities and penalties when safeguards fail.
For users, the meaningful test is dependable performance, clear limitations and accountability for mistakes. Governments face a broader challenge: encouraging investment without allowing optimism to substitute for evaluation, while ensuring that productivity gains benefit society rather than concentrating among a handful of companies.
Trump can change the government’s vocabulary. An executive order cannot establish scientific superiority. That requires evidence, independent testing and reproducible results. Between “artificial” and “super,” the essential question remains whether we are building institutions capable of making increasingly powerful technology worthy of public trust.