The September 2026 White House order makes “Super Intelligence” newly important in U.S. government communication, but it should not be described as the moment the concept of superintelligence was invented. Long before this order, researchers, authors and AI labs were already discussing systems that could exceed human intellectual performance and the governance problems such systems might create.
A clear recent example is OpenAI’s 2023 Superalignment initiative. The company described superintelligence as AI much smarter than humans and argued that existing alignment techniques might fail when people can no longer reliably supervise a stronger system. Whatever one thinks about the timeline or probability of that scenario, the public document demonstrates that “superintelligence” was established technical language years before the 2026 federal terminology shift.
What changed in September 2026 is institutional usage. The executive order directs federal executive agencies to use SI instead of AI in official non-statutory communications and asks for proposed legislative language that could create a federal SI definition. That is a policy decision about naming and future legal structure, not evidence that every current AI system has suddenly crossed a new technical capability threshold.
This distinction matters for historical accuracy and for marketing. A headline saying “AI has been renamed SI everywhere” would overstate the order. A more accurate headline is that the U.S. executive branch has adopted SI as its preferred terminology while the wider technology community continues to use several capability terms.
As language evolves, good communication should preserve the difference between a government naming policy, a scientific concept and an actual measured capability. Those three layers may influence each other, but they are not automatically the same thing.
Technology can expand the range. Taste, context and craft still decide what should remain.