Humanitarian-disarmament treaty for a class of weaponised AI

Build a long-running coalition of NGOs, sympathetic states, faith communities, and survivors that campaigns for a binding multilateral treaty banning or restricting a specific class of AI weapon system. The coalition runs in parallel through the formal UN process (CCW, General Assembly, regional forums) and through public mobilisation in client states, while assembling the evidence base — humanitarian impact, military doctrine, weapons-review precedent — that the diplomatic track will draw on.

An actor chooses this approach because nation-state behaviour around weapons is bound by treaty more than by anything else, and because the humanitarian-disarmament lineage has a working playbook for it: landmines, cluster munitions, blinding lasers. The form is slow but it does not require any one government — once a critical mass of middle powers commits, the treaty creates a norm that constrains even non-signatories, by changing what is procurable, fundable, and publishable.

It trades off pace for permanence. The horizon is a decade; the campaign carries dead weight through several political cycles before producing visible movement; and it depends on the great-power calculation that being seen to object is more costly than acquiescing — a calculation that can fail.

Verdict — good strategy, with sharp scope conditions

Strong where the conditions are met. The effects above are not adjacent to the strategy — they are its product: the 2023 First Committee resolution, the SG's 2026 deadline, the 2024 General Assembly resolution, the New York consultations, and the 42-state Brazil statement all sit on a single causal chain that the coalition orchestrated across thirteen years. Few civil-society strategies on AI have a comparable trail of attributable hard regulatory output. The coalition has also built the most durable cross-domain infrastructure the AI-policy field has produced — bridging disarmament veterans, faith communities, Nobel laureates, AI researchers, and parliamentary advocacy in seven-plus countries — which has survived the consensus block at the CCW (2014–2022) and the transition from HRW-housed coordination to a dedicated Geneva secretariat (2021).

The strategy is good only where the AI use case can be framed as a humanitarian crime — something that disproportionately harms civilians, lacks a clean military-utility justification, and can be banned categorically rather than regulated case-by-case. Autonomous weapons targeting humans clears that bar; algorithmic welfare scoring, generative-AI labour displacement, surveillance, and content moderation do not — they call for strategic litigation, coalition lobbying of binding regional regulation, or empirical audit and expose instead. Adopters who try to transpose the disarmament template to AI use cases without a humanitarian-crime frame produce coalitions that struggle to attract the faith and survivor communities that supply the strategy's moral weight.

Two further conditions earn the verdict. First, the class definition has to survive technical contestation. Mines and cluster munitions are physically definable; "weapons that target people without meaningful human control" is a software-and-doctrine definition that great powers will probe for ambiguity. The campaign has held its ground here, but the contested definition is the strategy's structural weak point and will be tested again in any negotiating text. Second, the strategy depends on middle-power state allies willing to spend diplomatic capital — Austria, Brazil, Mexico, Costa Rica, New Zealand, Ireland have done so consistently; a strategy without this bench is a strategy without a UN track.

Ecology

This strategy is fed by survivor-led testimony as evidence — the Automated by Design exhibition's filming in Nabi Saleh, Nairobi, and Rio de Janeiro is the campaign's deliberate import of survivor testimony into the multilateral frame, the same lineage that anchored the Mine Ban Treaty. It is reinforced by counter-narrative framing — "meaningful human control," "digital dehumanisation," and #VoteAgainstTheMachine are the framings that move the file through press cycles and into UN documents. It draws moral force from faith-institutional moral framing of AI — the Dalai Lama's 2014 Nobel Laureates declaration, Pope Francis's 2020 and 2024 interventions, the Holy See's diplomatic engagement at CCW. It is reinforced by open-letter collective signatory action — the Buenos Aires AI-researcher open letter (2015, 1,000+ signatories including Hawking, Wozniak, Russell) supplied the scientific authority the diplomatic track needed.

It shares moral and tactical lineage with mass protest for an AI moratorium and tech-worker refusal inside AI labs — Project Maven sat across all three, with Google-employee organising as the proximate driver and the disarmament framing as the discursive context. Newer AI-policy organisations — Encode Justice's US federal restrictions on AI in nuclear command-and-control among them — explicitly cite the autonomous-weapons strategy as part of their inheritance, which is the surest sign that even on its slow timeline it has become a working template the rest of the AI-good movement is drawing on.