On 8 August 2024, in New York, the delegations negotiating the UN Convention against Cybercrime voted on a single paragraph. It allows a state to decline a request for assistance if it has substantial grounds to believe the request is aimed at persecution. That means prosecuting or punishing someone because of their sex, race, language, religion, nationality, ethnic origin or political opinions. At Iran's request, the Committee voted on deleting it. The deletion failed by 109 votes to 25. The states that kept the clause came from every region: Brazil, Kenya and the Philippines voted alongside all 27 EU Member States. Proposals to delete the Convention's human rights clause in Article 6(2) and its safeguards article, Article 24, failed the same day.
While a victory for human rights, it also marked the limits of what written text can accomplish on its own. The persecution clause, like all refusal grounds in the treaty, gives states the option to say no, but it does not force them to do so. Where a duty to refuse exists, it comes from a state's existing human rights obligations, which Article 6(1) requires it to respect when applying the treaty. In short, the negotiators secured a permission, not a protection.
Since then, Europe's debate about the Convention has tended to focus on the text: whether to sign, which reservations to oppose, how serious its human rights gaps are. That is now the wrong fight. As of 2 November 2026, the Convention has 96 signatories. Three states (Qatar, Azerbaijan and Viet Nam) have become parties, of the forty needed for entry into force, and the Council has approved it on the EU's behalf. The question that matters now is how states will apply it. The practical meaning of the Convention's safeguards will be shaped by how the first states to apply them interpret them. Europe cannot control how others will use that discretion. It can decide how it uses its own. EU Member States should declare, on ratification, how they will refuse requests under the Convention, and publish figures on their refusals every year. Until they do, Europe has little standing to ask anyone else for reasoned refusals.
A Framework of Discretions
Read closely, the Convention is neither the disaster its critics fear nor the shield its defenders promise. It is a set of discretions.
The critics have a point, and the Convention's defenders should concede it. The four-year threshold for "serious crime" opens the door to cross-border evidence gathering well beyond the Convention's eleven offences. It is too low. As ARTICLE 19 has argued, any state can meet it by creating an offence with a heightened domestic penalty. The domestic procedural powers reach further still, to electronic evidence of any criminal offence. Article 24 requires proportionality and, as appropriate, judicial or other independent oversight and an effective remedy, but leaves the detail to domestic law. Similar concerns were set out on this blog by Samaya Anjum, Jason Pielemeier and Elonnai Hickok.
But the text also gives a careful state ample room to refuse. Where the Convention's own assistance rules apply, as they will between states with no other assistance treaty between them, a requested state may refuse for lack of dual criminality (Article 40(8)) or under the persecution clause (Article 40(22)). Article 40(21) adds four general grounds. The broadest allows refusal where the requested state considers a request likely to prejudice its sovereignty, security, ordre public or other essential interests, a judgement that is the state's own to make. Behind all of this stands Article 6(2), which bars any reading of the Convention as permitting the suppression of human rights.
The same rules impose two duties. Reasons must be given for any refusal, and the requested state must consult before refusing on the general grounds or postponing. These are the real safeguards, and they are thin. Consultation is not required before refusing under the persecution clause or for lack of dual criminality, and the Convention does not say what a refusal's reasons must contain. Whoever applies these provisions first, and consistently, will in effect write their commentary. If the first users refuse opaquely, or assist without asking hard questions, that practice risks becoming the norm. As one practitioner put it, drawing on the organised crime convention, states "must be clear on their conditions for cooperating or not".
The Gap Is Predictability, Not Law
Much of the legal groundwork has already been laid. The Council of Europe's own comparison found the core offences and procedural powers of the UN draft more or less identical to those of the Budapest Convention. That treaty turns 25 this month and has just gained its 84th Party in Kazakhstan. Both treaties apply their own assistance procedures mainly where no other assistance treaty binds the parties concerned. There are 109 UN Member States outside the Budapest Convention, many of them in the Global South, where networks of bilateral assistance treaties are often thin. For them, the UN Convention offers something new: a common set of electronic evidence powers.
Neither treaty, on its own, offers predictability. In 2014 the Budapest Convention Committee found that response times of six to 24 months "appear to be the norm" in mutual legal assistance, and that detailed statistics were not available. In 2024, the length of the procedure was still the most significant challenge for EU judicial authorities seeking electronic evidence from outside the Union, cited by 73 per cent of respondents. When a request goes out, a prosecutor cannot know on what conditions the answer will come, or whether it will arrive in time to matter. More debate about the text will not change that. Practice will.
Europe Learned This the Hard Way
Europe has a better claim than most to know what works, because it has had to solve a comparable problem among its own members. In 1999 the Tampere European Council endorsed mutual recognition. Since then, the Union has narrowed the grounds on which Member States may refuse one another's decisions. The European Arrest Warrant, for instance, removed the dual criminality check for 32 categories of offence. The trust this required, as one review of those twenty years observed, "cannot be established by decree".
What emerged was not the absence of refusal but its discipline. When national courts confronted serious doubts about detention conditions and judicial independence elsewhere in the Union, the Court of Justice did not open a general exception. In Aranyosi and Căldăraru and LM it set a two-step test. First, is there a real risk arising from systemic or generalised deficiencies? Second, would the person concerned be exposed to it? Before the surrender procedure can be brought to an end, the executing authority must request supplementary information and postpone its decision. Later judgments held that systemic deficiencies alone do not justify refusal. The result is a narrow, reasoned and reviewable exception. It let the system absorb grave doubts about individual cases without reopening the principle of mutual recognition.
The same logic runs through Europe's most direct cybercrime instrument. Since 18 August 2026, European Production Orders have let judicial authorities require service providers offering services in the Union to hand over electronic evidence directly. Even so, where traffic or content data are sought, the enforcing authority must usually be notified. The one exception is where the offence has been, is being or is likely to be committed in the issuing state, and the person concerned resides there. The issuing authority must have reasonable grounds to believe both conditions are met. The enforcing authority must then, where appropriate, raise specified grounds for refusal, including, in exceptional situations, a manifest breach of a relevant fundamental right. Even Europe's most direct form of cooperation comes with a precisely defined way to say no.
The lesson is simple. Cooperation deepens when refusal is narrow, reasoned and open to review. It does not deepen when refusal is absent, and it does not deepen when refusal is a black box.
Europe's Own Record Is the Weak Point
Here Europe should be honest with itself. Seen from outside, the EU can itself be a demanding requested party. EU Member States may make assistance conditional on dual criminality or refuse it where that condition is not met. EU law also limits transfers of personal data to third countries for law enforcement purposes. They require an adequacy decision, appropriate safeguards or one of a few narrowly defined derogations. Those rules exist for good reasons. But seen from the other side of a request, a European refusal can look as opaque as anyone else's.
That is why the European experience cannot simply be exported. Mutual recognition grew inside a single legal order with a common court; nothing comparable exists among the Convention's parties. What can travel is the method: a stated test, applied with reasons, and consultation before a door is closed. And the method will only travel if Europe is seen to apply it to itself first.
It would find allies. The majority of August 2024 did not form around a European proposal; it formed around a safeguard that states from Brasília to Nairobi to Manila chose to keep. Coalitions shift by issue, however. In Vienna in January 2026, Russia and the African Group, represented by Egypt, led the campaign against open stakeholder participation in the future Conference of States Parties. Talks on the rules of procedure deadlocked over that question, as Faiz Naeem has explained on this blog. Europe will win more of these fights by example than by instruction.
What Europe Should Do
The central demand: declare how you will refuse, and publish the results.
On ratification, each EU Member State should adopt and publish a national policy, coordinated with the others, setting out how its central authority will handle requests under the Convention. Such policies would not reinterpret the treaty. They would commit each state to practice that goes beyond it, describing its own conduct rather than prescribing anyone else's. A recent contribution to this blog has called on governments to put their reading of the safeguards on record; this goes a step further. In substance, the policies would contain four commitments. Where a request raises a risk of persecution or other serious human rights violations, the central authority would assess both the general situation in the requesting state and the circumstances of the person concerned before deciding whether to grant it, a two-step approach that mirrors the Court of Justice's test in Aranyosi and Căldăraru. It would consult the requesting state before refusing on any ground, including the persecution clause and the absence of dual criminality, unless that would endanger the person concerned; Article 40(28) requires consultation only before refusals on the general grounds. Every refusal would identify the paragraph relied on and, as far as domestic law and the safety of the person concerned allow, its factual basis, going beyond the bare duty in Article 40(25) that reasons be given. And, as Article 40(7) strongly encourages, these procedures would apply even where a bilateral treaty would otherwise govern.
Then Member States should count. For each incoming request, central authorities should record the requesting state, the legal basis, the measure sought, the dates of receipt and execution, and the outcome. For any refusal, they should note the ground relied on and whether consultation took place. Published as annual aggregates, these figures would let partners hold Europe to its word. They would also give the future Conference a template for the information it may require under Article 57(6). Preservation requests are the obvious place to start: Article 42(4) excludes dual criminality as a condition for preservation, subject to a possible reservation, and speed matters most there.
Three further steps would make that commitment credible.
Ratify now.
No EU Member State has yet ratified, and the EU can deposit its own instrument only once at least one Member State has done so. Ratification also matters for the objections to incompatible reservations that the Council's decision on the Convention foresees. Under the International Law Commission's Guide to Practice on Reservations, signatories may formulate objections, but these objections take effect only once the signatories are bound by the treaty.
Protect the safeguards where the rules are being written.
The European Commission's recommendation of 29 September 2026 already points the protocol talks, which open in Vienna in January 2027, in the right direction. It calls for an assessment of whether a protocol is needed before its substance is negotiated, for the meaningful participation of all relevant stakeholders, and for any new cooperation measures to remain subject to the Convention's grounds for refusal. Europe should add two specifics. On participation, it should not invent a new formula, but propose for the Conference of States Parties the one under which the Convention itself was negotiated, set by the General Assembly in 2021. On the protocol, it should seek a relationship clause like those in the three organised crime protocols. Article 61(4) already provides that a protocol is to be interpreted together with the Convention, and the Commission's directives refer to it. But joint interpretation does not by itself make the Convention's provisions apply to new offences. An explicit clause, stating that the Convention applies mutatis mutandis and that the protocol's offences count as offences established under it, would remove any doubt that Article 6, Article 24 and the grounds for refusal cover whatever the protocol adds.
Make cooperation easier to get right.
In the Kyoto Declaration, states undertook to endeavour to update tools such as SHERLOC, the Mutual Legal Assistance Request Writer Tool and the Directory of Competent National Authorities. According to its public description, the Directory does not yet cover the UN Convention. Article 54(10) asks parties to endeavour to make voluntary contributions to UNODC. Such contributions could close that gap before the Convention enters into force. A request that reaches the right authority, in an accepted language, is one that is easier to grant.
Saying No in the Open
The vote in New York kept the persecution clause in the text. Whether it ever protects anyone will be decided elsewhere: in ministries and prosecution offices still waiting for their first request under the new treaty. Europe can wait to see what others make of the Convention's discretion, and then complain about it. Or it can show, from the first day, what a good refusal looks like. Europe should declare how it will say no, and publish every year how it did.
The views in this paper are provided by the author in her personal capacity only.
