Wrapping Up the 2025 INTERPOL General Assembly in Morocco

The 2025 INTERPOL General Assembly (GA) in Morocco is over. By and large, it was a success. Of course, INTERPOL never admits that its annual meetings are anything but successes, but even by the standards of an outside observer – albeit one who attended the meeting as press – Morocco went about as well as it plausibly could have done.

There is, admittedly, unfinished business. The GA adopted changes to the Commission for the Control of INTERPOL’s Files’ (CCF) Statute that, like the 2024 Financial Statement and the CCF Activity Report for 2024, are not yet available. Nor has INTERPOL released – it likely never will – the vote totals in the six elections to INTERPOL’s Executive Committee. The GA may have adopted a resolution that promises greater transparency on INTERPOL governance, but evidently that transparency only goes so far.

The Morocco meeting in sunny Marrakech did feel somewhat more lightly attended than last year’s meeting in not at all sunny Glasgow. The same number of countries — 179 – were there, but only about 800 individuals made it to Marrakech, as opposed to 1,000 for Glasgow. The GA’s workload in 2025 – 14 resolutions – was also lower than 2024’s 18, and none of 2025’s resolutions drew more than token opposition. One of 2024’s resolutions was opposed by 38 countries: no resolution in 2025 was opposed by more than 2. All of 2025’s controversy appears to have been concentrated in those elections.

On the ground, the Moroccan authorities did a fine job, even if the announcement ahead of the GA that Morocco has strengthened its security cooperation with the PRC wasn’t comforting. Coffee – and a snack – in the media room would have been a nice touch, but you can’t deny the grandeur of the entryway into the main hall in which the GA held its meetings. Regrettably, INTERPOL did a lot less by way of media opportunities this time round. It did hold its first ever online briefing – on the Silver Notice pilot – which was a welcome initiative, though its technology did let it down slightly.

But there were none of the sessions with INTERPOL personnel that were a regular feature of 2024, and the INTERPOL Secretary-General’s press huddle bounced around the schedule like an unruly rubber ball. True, there were fewer press in Marrakech than Glasgow, so INTERPOL had less incentive to lay on briefings. But transparency is a two-way street: it’s hard to convince press to attend if you can’t offer them anything in advance – especially if you can’t even announce the precise location of the meeting more than 48 hours ahead of time.

Strangely, INTERPOL didn’t offer a full briefing on the capabilities it announced at the start of the GA. First, there is Nexus, which  INTERPOL describes as “a web-based messaging platform that uses artificial intelligence to help structure and deliver police information in a clearer, more accessible format, designed to help frontline officers receive critical intelligence faster.”

Second, there is AVA, “the first AI virtual assistant for law enforcement that operates entirely within INTERPOL’s protected cloud environment. It can rapidly process lengthy or complex material, translate instantly and summarize information to provide officers with actionable insights.”

It is true that one of the challenges for law enforcement today is the gap between the resources – especially electronic resources – available to better-off countries and those used by INTERPOL’s poorer members, and Nexus and AVA could be seen as efforts to close that gap. But both initiatives raise serious questions, even apart from the well-known tendency of AI models to be extraordinarily sensitive to prompts and to hallucinate answers.

If Nexus and AVA are processing information held in INTERPOL-maintained databases, this points out the challenges related to the vast amounts of information INTERPOL is now receiving – a subject addressed in the 2023 changes to the Rules on the Processing of Data. INTERPOL is supposed analyze this data but only return the results that comply with its rules – but if AI has the power to return summaries, it will be extraordinarily challenging to ensure that those summaries are “compliant.”

If, on the other hand, Nexus and AVA are processing information that remains in national databases, then INTERPOL runs a clear risk of violating Articles 2(1) and 3 of its Constitution by aiding – through the provision of its AI tools – the commission of acts at the national level that are incompatible with the UDHR or that take INTERPOL beyond ordinary law crimes.

Nexus and AVA raise other questions. If they are versions of commercial AI models, INTERPOL would be relying on tools trained on data that have nothing to do with police work – and this might be neither effective or safe. If they are in-house AI tools, they must have been trained off INTERPOL-held data provided by its national members, which raises questions about the reliability of the data (and hence the tool), as well as whether INTERPOL’s democratic member nations should consent to the use of their data to develop tools for the benefit of autocracies.

Finally, there is the familiar problem of the sustainability of INTERPOL’s initiatives and its tendency to chase trends – a tendency related to both its desire to have appealing-sounding projects and to its funding model, which relies on project funding by member nations to supplement the dues it collects. The result is projects like INTERPOL’s ID-Art app – launched with fanfare in 2021, not updated for the past year, and now languishing with 8 ratings in Apple’s App Store. AI is obviously hot now, and Nexus and AVA sound uncomfortably like more publicity-seeking projects. Keeping these tools up to date will at a minimum be an ongoing and expensive technical challenge.

The questions around the compliance of INTERPOL’s new AI tools with its Constitution is the kind of challenge that the CCF will increasingly be called upon to address. There are serious issues on the horizon – from the ongoing review of the CCF Statute to the GA’s decision to emphasize the importance of Article 2(1) to the legal framework for INTERPOL’s data processing (which could presage a change to INTERPOL’s treatment of refugees).

The Committee on the Processing of Data, which is leading the CCF Statute review, has rightly sought public contributions to aspects of this review, a practice it will hopefully continue. But it is now all the more important that the nations that want INTERPOL to work within its Constitutional boundaries – which are essential to its legitimacy and its safety from lawsuits – start thinking well in advance about their candidates and strategy for the CCF elections in 2026, in which all seven CCF seats will be open.

The fact that the 2026 General Assembly meeting will be held in Hong Kong adds a good deal of poignancy to the upcoming CCF elections. INTERPOL is always only one GA away from bad decisions. It largely dodged that bullet this year. But there is always next year.

—

Image: Shutterstock