# Holding a plasma in shape.

Researchers used reinforcement learning to control magnetic confinement in a research tokamak.

Canonical: https://brightaifuture.com/discoveries/fusion
Format: story
Source publication: 2022-02-16
Bright publication: 2026-09-05
Substantive update: None recorded
Evidence and review: Demonstrated; confidence: unassessed; source-checked; legacy source check. Legacy source check; no named human reviewer is recorded in this projection.

## The human problem

Hot plasma must be controlled within a complex magnetic system.

## The prior constraint

Controllers require specialized engineering for target configurations.

## AI’s actual role

A trained policy commanded the tokamak’s magnetic coils.

## The documented result

The controller demonstrated multiple plasma configurations on TCV.

## Why it may matter

A tool for researchers exploring fusion reactor designs.

## Limitations

Plasma control is not net energy production or commercial fusion.

## Unresolved questions

Test broader conditions and more demanding devices.

## Provenance and history

{
  "dates": {
    "eventDate": null,
    "publicationDate": "2022-02-16",
    "captureDate": null,
    "lastReviewedDate": "2026-09-05"
  },
  "provenance": {
    "origin": "legacy-projection",
    "externalId": "fusion"
  },
  "revisions": [],
  "corrections": []
}

## Original sources

- [Magnetic control of tokamak plasmas through deep reinforcement learning](https://www.nature.com/articles/s41586-021-04301-9)

## Continue exploring

- [Concise evidence record](https://brightaifuture.com/discoveries/fusion)