Most people have never heard of the General-Purpose AI Code of Practice. Yet it now shapes how the AI models behind everyday tools — the chatbots, writing assistants and coding helpers used in offices across Ireland — are documented, tested and explained in Europe.

On 31 July 2026, the European Commission published an updated list of more than 180 signatories to the Code. Two days later, on 2 August, the Commission's power to enforce the EU AI Act's rules for general-purpose AI models came into application.

That combination has turned the Code from a policy document into one of the most important working parts of Europe's AI rulebook. Here is what it is, who has signed it, and why it matters to Irish businesses that never build a model themselves.

What the Code of Practice Is

The EU AI Act sets obligations for providers of general-purpose AI models — large, flexible systems that can be used for many different tasks and built into countless products. Those obligations have applied since 2 August 2025.

But the Act itself is written in broad legal terms. Providers needed a practical way to show they were meeting it. The Code of Practice fills that gap. According to the Commission, providers can demonstrate compliance by relying on the Code, which was prepared by independent experts through a multi-stakeholder process.

That process was substantial. The Commission's AI Office facilitated it, independent experts chaired it, and close to 1,000 stakeholders took part, alongside representatives of EU member states and European and international observers. The finished Code was endorsed by the Commission and by the AI Board, which is made up of member states.

The Code is organised around three themes. Transparency covers the documentation providers keep and share with the businesses that build on their models. Copyright covers how providers respect EU copyright law. Safety and security applies to the most advanced models, and deals with how serious risks are assessed and managed.

Who Has Signed — and Why It Matters

The list of signatories reads like a directory of the modern AI industry. It includes Amazon, Anthropic, Google, IBM, Microsoft, Mistral AI and OpenAI, as well as companies such as Cohere, ServiceNow and WRITER.

European developers are well represented too. Aleph Alpha, Black Forest Labs, Almawave, Fastweb and LINAGORA are among the names on the list, alongside a growing number of smaller firms. By the end of July, the total had passed 180.

That level of take-up is a strong vote of confidence. It suggests that most of the companies building the models Europeans rely on see the Code as a workable, sensible way to meet their obligations — and would rather work inside a shared framework than argue against it.

It also creates a level playing field. When the leading providers are signed up to the same commitments, smaller developers have a clear benchmark, and businesses buying AI services can compare suppliers on a common basis.

A Living Rulebook: The Signatory Taskforce

One of the smartest features of the Code is that it is not frozen in time. AI models improve quickly, and the risks they pose change with them. A rulebook written once and left alone would soon fall behind.

To deal with that, signatories have set up a Signatory Taskforce, chaired by the Commission's AI Office. It gives providers and the regulator a standing forum to work through how the Code is applied as the technology develops, and to share what is working.

This is a very European way of doing things: clear rules in law, backed by a practical, collaborative mechanism for applying them. It keeps the regulator close to the technology and gives companies an early view of what is expected, which reduces uncertainty on both sides.

Ahead of enforcement in August 2026, the Commission's AI Office also worked with signatories on how the Code should operate in practice, so that the move from voluntary commitment to enforceable obligation would be as smooth as possible.

What It Means for Irish Businesses

For most Irish companies, the Code is less about their own obligations and more about the tools they buy. If a business uses a chatbot, a writing assistant or an AI feature inside its software, there is a good chance the model underneath comes from a Code signatory.

That brings real benefits. Signatories commit to sharing documentation with the businesses that build on their models, which makes it easier for Irish firms to understand what a model can and cannot do, and to meet their own responsibilities under the AI Act.

It also gives buyers a useful question to ask. When choosing an AI supplier, checking whether the underlying model provider has signed the Code is a quick, meaningful signal of how seriously that provider takes transparency, copyright and safety.

And there is local support. Ireland's Regulation of Artificial Intelligence Act 2026, in force since 31 July, established the AI Office of Ireland as the national single point of contact for the EU AI Act, with a stated role in promoting AI literacy and public awareness. Irish organisations looking for guidance now have a clear place to start.

Why Voluntary Codes Work

It might seem strange that one of the most important parts of a binding law relies on a voluntary document. In practice, codes of practice are a well-established way of turning broad legal duties into clear, day-to-day steps that companies can follow.

The logic is simple. Laws are written to last, so they set out principles rather than technical detail. Codes can be more specific, and they can be updated as the technology and the risks change. For something moving as fast as AI, that flexibility is essential.

For providers, signing the Code offers a clear route to showing they meet the rules. Rather than each company working out its own interpretation of the AI Act and hoping the regulator agrees, they can follow a shared set of commitments that the Commission and member states have already endorsed.

For the regulator, broad take-up makes supervision more consistent. When most major providers are working to the same framework, it becomes much easier to compare approaches, spot gaps and share good practice across the industry.

And for everyone else, it means the most widely used AI models in Europe are being documented, explained and checked against a common standard. That is a quiet but significant step towards the kind of trustworthy AI that businesses and the public both want.

The strong take-up also sends a clear message beyond Europe. When the companies building the world's leading AI models choose to sign a European code, it shows that thoughtful rules and fast-moving innovation can sit side by side. Other regions watching how to govern AI now have a working example to learn from, built on cooperation rather than confrontation, and backed by clear law.

The Bottom Line

The General-Purpose AI Code of Practice has quietly become one of the most successful parts of Europe's AI rulebook. Built with close to 1,000 stakeholders, endorsed by every member state and signed by more than 180 providers — including the biggest names in AI — it gives the industry a practical, shared way to meet the law, and gives businesses a clearer picture of the tools they rely on. With enforcement now in place and a taskforce keeping the Code up to date, Irish companies have a simple, useful step available today: know which models sit behind your AI tools, and check that their providers have signed up.