Council of Europe AI Treaty (CETS 225)

The Council of Europe AI Treaty is the Council of Europe Framework Convention on Artificial Intelligence and Human Rights, Democracy and the Rule of Law, CETS No. 225. The UK signed it on 5 September 2024, and the UK government describes it as the first legally-binding international treaty on AI. This page covers what’s independently confirmed about the treaty’s existence, formal name, and signing — it deliberately does not state the treaty’s current ratification or entry-into-force status, which we could not confirm from an independent source at time of writing (see the note below).

Formal name: Council of Europe Framework Convention on Artificial Intelligence and Human Rights, Democracy and the Rule of Law
Treaty number: CETS No. 225
Opened for signature / UK signed: 5 September 2024
Signed for the UK by: Lord Chancellor Shabana Mahmood
Non-Council-of-Europe states invited to sign: United States, Australia
UK government’s own description: “the first legally-binding international treaty on AI”
Ratification / entry-into-force status: not stated on this page — not independently confirmed (see below)

What CETS 225 actually is

The treaty’s formal title, per the UK Foreign, Commonwealth & Development Office’s own Treaty Action Bulletin for September 2024, is the Council of Europe Framework Convention on Artificial Intelligence and Human Rights, Democracy and the Rule of Law. It’s catalogued in the Council of Europe’s treaty series as CETS No. 225. The bulletin lists the UK’s signature of the treaty under its “Multilateral treaties: UK actions” section, alongside a reference to the Council of Europe’s own conventions database.

A “framework convention” is, by name, a treaty that sets out principles and obligations at a general level rather than prescribing detailed technical rules — the Council of Europe has used the same structure before, for example in its Framework Convention for the Protection of National Minorities. That framing matters for how CETS 225’s obligations would likely translate into domestic law for any state that ratifies it, but this page does not attempt to summarize the treaty’s substantive text, since CASRAI could not access the Council of Europe’s own treaty documents directly (see the sourcing note below).

The signing: 5 September 2024

The UK signed CETS 225 on 5 September 2024, with the Ministry of Justice’s own announcement naming Lord Chancellor Shabana Mahmood as the UK’s signatory. The same announcement states that countries outside the Council of Europe are also being invited to become signatories, specifically naming the United States and Australia — a detail that matters because it signals the treaty’s drafters designed it to reach beyond Council of Europe membership, rather than as a purely regional instrument.

The Ministry of Justice’s announcement is explicit in its own framing of what the treaty is: it describes CETS 225 as “the first legally-binding international treaty on AI.” That’s the UK government’s characterization of the treaty’s legal character as a signed instrument — it is not, on its own, a statement that the treaty has entered into force for the UK or for any other state. Signature and entry into force are different steps in treaty law, and this page addresses only the first.

What this page does not confirm — and why

CASRAI could not independently confirm CETS 225’s ratification status or entry-into-force date for this page. Direct attempts to reach the Council of Europe’s own treaty pages (coe.int and its treaty-office mirror at rm.coe.int) failed, as did a Wayback Machine lookup for an archived copy. A follow-up search of GOV.UK’s own publication index for a ratification-specific bulletin returned nothing relevant. Rather than infer a status from silence, this page states plainly what it does and doesn’t cover: the treaty’s existence, its formal name, and its signing by the UK on 5 September 2024 — not whether it is currently in force.

That gap is worth flagging rather than glossing over, because “signed” and “in force” mean different things for a multilateral treaty of this kind: a framework convention typically requires a minimum number of ratifications, including from a set number of Council of Europe member states, before it takes legal effect for anyone — and separately again for each individual state that ratifies it afterward. CASRAI will update this page once a working, non-coe.int-dependent primary source for that status is found.

How CETS 225 fits next to the binding AI statutes already in force

CETS 225 sits at a different level than the national and regional AI statutes CASRAI already covers. The EU AI Act’s GPAI Code of Practice, California’s SB 53, and South Korea’s AI Basic Act are each binding within a single jurisdiction — an EU regulation, a US state statute, and a national law, respectively. CETS 225, by contrast, is a multilateral treaty open to signature by states well outside the Council of Europe’s own membership, including the US and Australia per the UK’s own announcement above. If enough states ratify it, a framework convention of this kind is designed to have a broader geographic reach than any single national AI statute — but that reach depends entirely on ratification, which is exactly the part this page cannot yet confirm. See CASRAI’s Frontier AI Safety Timeline for how CETS 225’s September 2024 signing lines up against the EU AI Act, SB 53, the RAISE Act, and the other statutes already in force.

The NIKOLAI angle: a gap, not yet a mapping

NIKOLAI is CASRAI’s own frontier-AI-safety dictionary — an independent, unendorsed reference work, not an official record of any lab’s, regulator’s, or treaty body’s terminology. NIKOLAI says so about itself, in its own words: “No lab, evaluator or regulator has endorsed NIKOLAI, reviewed a mapping, or been consulted on this dictionary.” That’s the genuine point of contrast worth drawing here. NIKOLAI’s own vocabulary is deliberately non-binding — a voluntary reference, not a standard. CETS 225, once in force, is the opposite kind of instrument: actual treaty law between states, not a declaration any party is free to disregard.

NIKOLAI’s Track N9, Commitments and Governance, is where the dictionary catalogues exactly the kind of review-and-amendment mechanics that a multilateral framework convention would also need. Its Framework Update Protocol element defines the concept as “an if-then rule set governing when and how a safety framework may be revised — the declared review cadence, the triggering conditions for an out-of-cycle review, who may propose a change, who must approve it, the publication deadline after approval, and whether a change-log is required.” We checked that element page directly before writing this: its crosswalk rows currently cover Anthropic’s Responsible Scaling Policy, the EU AI Act’s GPAI Code of Practice, and California SB 53 — three voluntary or statutory frameworks, each issued by a single lab, bloc, or state. No crosswalk row exists yet for CETS 225.

That’s a sequencing gap, not an oversight: NIKOLAI’s crosswalk rows are built from published organizational text, and neither CETS 225’s own amendment provisions nor its ratification status could be independently confirmed for this page. Once a working, coe.int-independent source for the treaty’s actual text is found, whether CETS 225 has a revision mechanism comparable to Framework Update Protocol becomes a fair question to ask — and, if it does, a legitimate candidate for a future NIKOLAI crosswalk row. We’re flagging that here, not asserting it.

Frequently asked questions

What is the Council of Europe AI Treaty officially called?

Its formal title is the Council of Europe Framework Convention on Artificial Intelligence and Human Rights, Democracy and the Rule of Law, catalogued as CETS No. 225.

When did the UK sign the treaty?

5 September 2024, per both the FCDO’s Treaty Action Bulletin for that month and the Ministry of Justice’s own announcement.

Who signed the treaty on the UK’s behalf?

Lord Chancellor Shabana Mahmood, per the Ministry of Justice’s announcement.

Is the United States a party to the treaty?

Not confirmed as a signatory by the sources this page relies on. The UK government’s announcement states the US and Australia were invited to become signatories as countries outside the Council of Europe, but that is an invitation, not confirmation that either country has signed.

Has CETS 225 entered into force?

This page does not state that, deliberately. CASRAI could not independently confirm the treaty’s ratification or entry-into-force status from a source it could verify outside the Council of Europe’s own site, which was not accessible at time of writing.

Is this the same thing as the EU AI Act?

No. The EU AI Act is a regulation of the European Union, binding on EU member states through EU law. CETS 225 is a Council of Europe treaty, open to a wider set of signatories including non-European states, and operates under international treaty law rather than EU law.

Sources

UK Foreign, Commonwealth & Development Office, “Treaty Action Bulletin: September 2024,” published 1 October 2024 — formal treaty title and CETS No. 225 designation, listed under UK signature actions.
UK Ministry of Justice, “UK signs first international treaty addressing risks of artificial intelligence” — 5 September 2024 signing date, Lord Chancellor Shabana Mahmood as UK signatory, the “first legally-binding international treaty on AI” description, and the US/Australia invited-signatory detail.

CASRAI attempted to verify the treaty’s ratification status directly against the Council of Europe’s own site (coe.int and rm.coe.int) and via the Wayback Machine; all three attempts failed or were refused at time of writing. This page will be updated once an accessible primary source for that status is found.

The Council of Europe AI Treaty (CETS 225): What’s Confirmed So Far