---
title: "Super Intelligence (SI)"
description: "Since Executive Order 14434, U.S. agencies call AI \"Super Intelligence\" or \"SI.\""
url: "https://sonomos.ai/glossary/super-intelligence/"
source: "https://sonomos.ai/glossary/super-intelligence/"
---

# Super Intelligence (SI)

The U.S. executive branch's official name for artificial intelligence since September 2026, and, separately, the research term for a hypothetical AI that outperforms humans in virtually every domain.

## In depth

"SI" now carries two meanings that describe very different things. In U.S. federal usage, Executive Order 14434, signed on September 29, 2026, directs agencies to use "Super Intelligence" and "SI" in place of "Artificial Intelligence" and "AI" in official correspondence, public communications, websites, reports and other non-statutory documents. The order renames the technology; it does not create a new category of it. Until Congress or a later Presidential action says otherwise, SI covers what the statutory definition of artificial intelligence in 15 U.S.C. § 9401(3) covers: the same chat assistants, coding agents and models people already call AI. The order does not alter previously issued regulations, and it directs federal agencies rather than private companies. In AI research, superintelligence (often artificial superintelligence, or ASI) is a much higher bar. Nick Bostrom's 2014 definition is "any intellect that greatly exceeds the cognitive performance of humans in virtually all domains of interest." No system is generally accepted as meeting it, though several labs say they are building toward it. For privacy and compliance the rename changes nothing of substance: HIPAA, the GDPR, the EU AI Act, state privacy laws and professional-conduct rules attach to what a tool does with data, not to its name. The capability trend still matters. Research presented at ICLR 2024 (Staab et al., "Beyond Memorization") found that large language models inferred personal attributes such as location, income and sex from ordinary text with up to 85% top-1 accuracy, so the more capable the model, the more it can learn from whatever context a prompt still contains.

## Examples

- A federal contractor receiving agency guidance about "SI tools": under the order's definition that means ChatGPT, Claude, Gemini, Copilot and the rest, not a new class of system.
- A law firm whose AI-use policy says "AI": the policy still covers the same tools when a client or agency calls them SI, though adding "SI" as a defined synonym heads off the argument.
- A research paper on artificial superintelligence (ASI): it describes a hypothetical future system, not the assistant in your browser today.

## How Locke handles super intelligence (si)

Locke works on the data, not the label. It screens what you send to AI services (SI services, if that is what your agency or client calls them) and masks sensitive data on your device before the request leaves. Detection does not depend on which model receives the prompt, so a more capable model changes nothing about what Locke removes first; it only raises the stakes of anything left in.

[See the product](https://sonomos.ai/locke/)

## Related terms

- Privacy Layer for AI

  A control point between a user and an AI service that detects, transforms, or blocks sensitive data before it reaches the model.
- EU AI Act

  The European Union's comprehensive AI regulation, which classifies AI systems by risk level and imposes conformity assessments, transparency obligations, and prohibitions on certain high-risk uses.
- Data Minimization

  The privacy principle of collecting and processing only the minimum personal data necessary for a specific, stated purpose.