# General-Purpose AI Code of Practice

> Steps AI model makers may sign up to. They help show the EU AI Act is met.

**Field:** Artificial Intelligence  
**Category:** Ethics & Regulation  
**Also known as:** GPAI Code of Practice, AI Act Code of Practice, Code of Practice for General-Purpose AI

## Think of it like

It is like a worked answer sheet. Follow it, and the teacher can see you did the sum right.

## Definition

The General-Purpose AI Code of Practice is a set of written steps that helps model makers follow the EU AI Act. It came out on 10 July 2025, after several drafts written with experts. It has three chapters: openness, copyright, and safety and security. The last one is only for the biggest models. Signing it is a choice, not a duty. A maker that signs it and sticks to it has an easier way to show it meets the law.

## Example

Anthropic, Google, Microsoft and OpenAI signed all of it. xAI signed one part. Meta did not sign.

## Why it matters

The law says what makers must do, but not how. The code fills in the how, so makers and the AI Office share one yardstick.

## Common mistake

Thinking a maker that does not sign is breaking the law. It still has to meet the Act, in its own way, and may face more questions.

## Three principles

- **Three chapters.** Transparency, copyright, safety and security.
- **A choice.** Signing it is not a duty.
- **Easier proof.** Following it shows you meet the law.

## Related terms

- [General-Purpose AI Model (GPAI)](https://learn.benibix.com/term/ai/general-purpose-ai-model)
- [European AI Office](https://learn.benibix.com/term/ai/ai-office)
- [EU AI Act](https://learn.benibix.com/term/ai/eu-ai-act)
- [Model Card](https://learn.benibix.com/term/ai/model-card)
- [Training Data Copyright](https://learn.benibix.com/term/ai/training-data-copyright)

![A row of four boxes joined by arrows: a model maker of general-purpose AI, a provider who builds the system and puts its name on it, a deployer who uses it at work, and the people the AI makes calls on. Above, the AI Office checks the model maker, and a national authority in each EU country checks the provider and the deployer. Under the model maker sits the optional Code of Practice. A bracket under the provider and the deployer says both must train their staff in AI literacy.](https://learn.benibix.com/diagrams/ai4-ai-act-roles.svg)

---

Source: https://learn.benibix.com/term/ai/gpai-code-of-practice
Author: Behnia Zarei (a developer in Vienna, Austria, who builds and teaches data, AI and business intelligence at BENIBIX)
Publisher: BENIBIX Learn (https://learn.benibix.com/)
Last updated: 2026-11-18
Licence: CC BY 4.0 (https://creativecommons.org/licenses/by/4.0/). Quote freely with attribution.
