Artificial intelligence (AI) is the broad category: a machine-based system that makes predictions, recommendations, or decisions for a set of objectives. Machine learning is one way to build such a system, in which a program adjusts itself using data instead of relying only on rules a developer wrote by hand. Every machine learning product counts as AI, but under the definitions used by U.S. law and NIST a system does not have to learn to qualify.
Key takeaways
- U.S. law defines AI as a machine-based system that makes predictions, recommendations, or decisions for human-defined objectives. Learning is not part of that definition.
- NIST describes machine learning as computer systems that adapt and learn from data to improve accuracy.
- The AI4K12 guidelines, funded by the National Science Foundation, treat learning as one of five big ideas in AI, next to perception, reasoning, natural interaction, and societal impact.
- A product labeled "AI-powered" tells a school little until the vendor says whether the system learns from data and which data.
The definitions that regulators and agencies use
The United States has a statutory definition. It sits in 15 U.S.C. 9401(3) and is quoted in a footnote of the Department of Education's final priority on advancing AI in education, published in the Federal Register in April 2026. It describes AI as "a machine-based system that can, for a given set of human-defined objectives, make predictions, recommendations, or decisions influencing real or virtual environments."
NIST's AI Risk Management Framework 1.0, released in January 2023, adapts the same idea from the OECD and ISO/IEC 22989: an engineered or machine-based system that generates outputs such as predictions, recommendations, or decisions, designed to operate with varying levels of autonomy. Neither text mentions learning. A fixed set of if-then rules that recommends a course or flags a risk can fit.
Machine learning gets its own entry in the NIST glossary: "The development and use of computer systems that adapt and learn from data with the goal of improving accuracy" (from NIST SP 800-55 Volume 1). British Columbia's education ministry takes a different route in its K-12 guidance and describes AI as the capacity of machines to replicate human-like cognitive functions, including learning and problem-solving. Definitions differ in wording, so a school policy should state which one it adopts.
How the terms relate
The table is a study aid. The definition column quotes or closely follows the cited source. The example column is illustrative and does not describe any named product.
| Term | Meaning in plain words | Illustrative classroom example | Definition drawn from |
|---|---|---|---|
| Artificial intelligence | A machine-based system producing predictions, recommendations, or decisions for set objectives | Software that suggests the next practice set for a student | 15 U.S.C. 9401(3); NIST AI RMF 1.0 |
| Rule-based system | AI whose behavior comes from rules written in advance, with no training on data | A checker that applies a fixed list of spelling rules | Descriptive label, not a quoted definition |
| Machine learning | Systems that adapt and learn from data to improve accuracy | A tool that estimates which skills a student has mastered from thousands of past answers | NIST glossary (SP 800-55 Vol. 1) |
| Generative AI | Models that imitate the structure of their input data to produce new text, images, audio, or video | A chatbot that drafts feedback on a paragraph | NIST AI 600-1 (July 2024), quoting an executive order |
How the K-12 AI guidelines place machine learning
AI4K12 is an initiative jointly sponsored by the Association for the Advancement of Artificial Intelligence (AAAI) and the Computer Science Teachers Association (CSTA), with funding from the National Science Foundation under grant DRL-1846073. It organizes AI education around five big ideas: perception, representation and reasoning, learning, natural interaction, and societal impact.
The learning idea is stated in three plain sentences. Computers can learn from data. Machine learning is a kind of statistical inference that finds patterns in data. The training data must usually be supplied by people, though a machine sometimes acquires it on its own, and tremendous amounts of data are required for the approach to succeed. This framing explains why machine learning raises questions about data that a rule-based program does not.
Why the difference matters when a school evaluates a tool
A rule-based program behaves the same way tomorrow as today, and its logic can be read. A system trained on data inherits the strengths and gaps of that data, and its behavior can shift when it is retrained. The Department of Education's April 2026 priority defines AI literacy broadly: the knowledge, skills, and attitudes learners need to work with AI, build it, and judge what it can and cannot do well. Among its examples is teaching students to spot AI-generated disinformation online.
Questions that follow from the distinction:
- Does the product learn from data, or does it apply fixed rules?
- If it learns, whose data trained it, and does it keep learning from student input?
- What happens to accuracy for students whose work looks different from the training examples?
- Is the output a prediction, a recommendation, or generated content, and who checks it before it affects a student?
- When the vendor updates the model, how will the school learn about it?
These are discussion prompts for teachers and administrators, not legal or procurement advice. A school's privacy officer and counsel decide how rules apply to a specific purchase.
FAQ
Is machine learning a type of AI?
Yes. Machine learning is one method for building AI systems, the one in which the system adapts from data. AI also covers approaches that do not learn, such as systems driven by hand-written rules.
Is a chatbot such as ChatGPT AI or machine learning?
Both labels apply. A chatbot is a generative AI system, and generative models are trained on broad data, which is machine learning. NIST's generative AI profile of July 2024 uses the term for models that produce synthetic content such as text, images, and audio.
Does something need to learn to be called AI?
Not under the U.S. statutory definition or the NIST framework, which focus on producing predictions, recommendations, or decisions toward set objectives. Other documents, including British Columbia's guidance, mention learning as one of the human-like functions AI replicates.
When do students typically learn the difference?
AI4K12 publishes guidelines and progression charts across kindergarten through grade 12, and the Department of Education's 2026 priority supports age-appropriate AI and computer science education in K-12. Provincial and state curricula set the actual grade placement.
Related reading
- Traditional AI methods compared with modern machine learning
- Support vector machines as a machine learning method
- How pattern recognition works in AI
- Using AI to support learning
More in AI in education.
Last verified: October 2026
Sources
- Artificial Intelligence Risk Management Framework (AI RMF 1.0), NIST AI 100-1, National Institute of Standards and Technology, January 2023
- Artificial Intelligence Risk Management Framework: Generative Artificial Intelligence Profile, NIST AI 600-1, July 2024
- Glossary entry "machine learning" (source: NIST SP 800-55 Volume 1), NIST Computer Security Resource Center
- Department of Education final priority on AI, with the statutory AI definition quoted in a footnote, Federal Register, April 13, 2026
- Big Idea 3: Learning, and the AI4K12 initiative overview, AI4K12 (AAAI and CSTA, National Science Foundation grant DRL-1846073)
- B.C. K-12 page on AI supports for schools, provincial education ministry, July 2024