Culture Key Translation · Signal analysis · Sources checked: 4 October 2026
The words we use for AI systems shape our expectations. When an official term changes, we need to examine exactly what changed: the language, the legal definition, the system’s capabilities or the conditions under which we trust it?
What happened
Executive Order 14434, dated 29 September 2026, directs US executive branch agencies to use the terms “Super Intelligence” and “SI” in place of “Artificial Intelligence” and “AI”, where permitted by law, in communications, websites, reports and other non-statutory documents.
The order uses the existing statutory definition of AI for its implementation. It also requires a proposed legislative definition to be submitted within 60 days. The text itself does not establish a new measurable capability threshold that a system must exceed to receive the label.
Primary source: Executive Order 14434, 29 September 2026.
The US National Institute of Standards and Technology already states on its public website that it is updating its communications in line with the order. This is an observable change in terminology. It is not a test of system capabilities.
Additional primary source: the institute’s public page.
Why it matters to people
A powerful label may influence how much we trust a system or how readily we accept its decisions. Here we identify a possible risk to examine. These sources do not provide evidence that the change in terminology caused a specific shift in public behaviour, procurement or the delegation of authority.
For the person using a system, the useful question remains concrete: for which task can they rely on it, under what conditions and within what limits?
What needs to stand behind a label
- A specific task: which capability was tested, and in what environment?
- A measurable criterion: what is performance compared against, and what result would weaken the claim?
- Limits and failures: which conditions were not tested, and where did the system fall short of expectations?
- Authorisation: what is the system allowed to do, regardless of its name?
- Human intervention: who can challenge a decision, and who can restrict its actions?
What we will watch
We will examine whether the forthcoming definition proposal includes testable criteria and how those relate to access permissions, responsibilities and restrictions. A test in one environment also needs to be distinguished from performance in actual use. These points remain questions for monitoring.
The Culture Key reading
The signal concerns the existing functions of Authority, Risk, State & decision logic, and Traceability & review. The bounded finding is that the prescribed terminology changed and there is a public example of its implementation. This does not establish increased capabilities or demonstrate that any particular use is safe.
Identity is a claim. Practice is evidence. Contradiction is the test. We apply the same question to every system: what observation supports the label, and what would make us reconsider it?
In your own life
If a tool you use is suddenly presented as “smarter”, before entrusting it with more tasks or access to your files you can ask: “What changed in practice, for which task was it tested and what limits remain?” Look for a specific result and the conditions of the test. This example is hypothetical and is not an established consequence of the order.
The question that remains
What evidence do we need before turning a new label into greater trust or authority?