Super intelligence, also spelled superintelligence and called artificial superintelligence (ASI), is a hypothetical form of AI that would exceed human ability across a broad range of intellectual tasks. A strong result on one task shows performance on that task; superintelligence describes ability across many domains.
In his September 22, 2026, UN address, President Trump called AI “Super Intelligence”. His broad public use of the phrase makes the narrower ASI concept worth explaining.
GC AI is enterprise legal AI software for in-house teams. As of September 2026, 2,100+ legal teams use GC AI, including teams at Hitachi, News Corp, and SKIMS.
A superintelligence claim requires evidence across domains. An in-house legal team can assess an available AI product by testing a representative legal task, checking the sources behind the answer, and identifying where counsel reviews or approves the work.
What Does Super Intelligence Mean?
Super intelligence describes a proposed level of capability across many domains. Three dimensions help make a claim testable: breadth (the kinds of tasks a system can perform), performance (how well it performs them), and autonomy (the steps it can take without direction). Google DeepMind’s Levels of AGI framework separates these dimensions.
A contract summary or science benchmark shows performance under its tested conditions. A broad superintelligence claim would also need results from different tasks and settings, along with the methods and limits behind those results.
AI, AGI, and ASI: The Basic Distinction
Term | Working Meaning | Question to Ask |
|---|---|---|
Artificial intelligence (AI) | A broad category of systems that perform tasks such as prediction, text generation, analysis, and decision support. | Which task and setting are we discussing? |
Artificial general intelligence (AGI) | A proposed level of AI with general-purpose ability across many kinds of intellectual tasks. | How broad is the demonstrated capability? |
Artificial superintelligence (ASI) | A hypothetical level of AI capability that would exceed human performance across a broad range of domains. | What evidence would support the claim across domains? |
Each term answers a different capability question. For procurement, legal teams also need to examine how a particular system handles their documents, protects confidentiality, and produces work counsel can verify.
Does Superintelligence Exist Today?
As of September 2026, artificial superintelligence remains hypothetical. OpenAI’s governance discussion treats it as a future capability. Strong results on selected tasks can be valuable. A claim of broadly superhuman ability would require evidence across tasks, settings, and conditions, with clear evaluation methods and limits.
What Could Superintelligence Do?
If such a system became possible, its broad abilities could support scientific discovery, complex planning, and problem solving across fields. These are future scenarios. Legal teams can assess available AI today through representative research, document analysis, and drafting tasks.
Is an AI Agent the Same Thing as Superintelligence?
An AI agent can carry out steps in a workflow, such as searching sources, comparing documents, or drafting an answer. Autonomy describes how much the system can do without a person directing each step. A superintelligence claim also concerns ability across many domains. DeepMind’s framework treats autonomy and capability as separate dimensions.
GC AI’s Chat 2.0 architecture uses specialized processes to research and analyze in parallel, then reconcile findings in one conversation. Counsel can evaluate the output, its sources, and the steps requiring approval against a specific legal task.
How Should In-House Legal Teams Evaluate a Superintelligence Claim?
When a vendor or executive uses the phrase, ask for the system, task, test conditions, and evidence. Those details turn a broad claim into an assessment of performance and oversight.
Which tasks, documents, jurisdictions, and conditions did the evaluation cover?
Who ran the test, what metric did they use, and what limits did they report?
Which steps can the system take on its own, and where does a person review or approve its work?
What information can the system access, retain, or share under the governing terms?
This is a practical reading of NIST’s AI Risk Management Framework, which asks organizations to map context, measure performance and risk, and define human oversight. The answers help counsel evaluate the system in the company’s workflow.
How Could Counsel Build a Source-Checked Brief?
Suppose the general counsel asks for a one-page explanation before a vendor meeting. Gather the vendor’s capability statement, its test methodology, the proposed data terms, and the primary technical definitions. Then record the claim each source supports.
GC AI’s Research feature can compare authoritative web sources and produce structured findings with citations. An illustrative request would be:
Compare the attached vendor materials with the provided AI, AGI, and ASI definitions. List each capability claim, the task and evidence offered for it, any stated limit, and the source passage. Separate tested behavior from forecasts. Flag questions about data access and human approval for counsel to resolve.
Counsel opens the citations, checks the vendor’s terms and methodology, and decides which questions to take into the meeting. Save the reviewed brief with links and a check date. AI for Legal Research goes deeper on verifying source-backed answers.
Start With the Work the System Will Do
Bring a recent legal research question to a GC AI demo. Define the jurisdiction, source types, and answer format, then compare the cited response with counsel’s own analysis. Record each supported conclusion, correction, and open issue. That record gives the team a basis for deciding when to use the workflow.
See It in Action: Research With Cited Sources
Frequently Asked Questions
Does “super intelligence” always mean artificial superintelligence?
The phrase can carry different meanings. President Trump used “Super Intelligence” as a broad label for AI in his September 22, 2026, UN address; artificial superintelligence (ASI) describes a hypothetical level of capability beyond human performance across many domains. Check the source’s definition before treating the terms as equivalent.
Is superintelligence possible?
It remains an open question. Researchers discuss it as a potential future capability, but results on specific tasks cannot establish whether AI will achieve broadly superhuman ability. A claim about its possibility should specify the technical assumptions and evidence.
When could superintelligence arrive?
There is no reliable arrival date. Forecasts depend on how researchers define the capability threshold and what progress they expect, and a benchmark result alone does not establish a timeline. In a legal briefing, label any date as a forecast and record its source and assumptions.
