Public-sector precautionary moratorium on AI

Movement organisations advocate that a specific public-sector body — a school district, a court system, a police department, a state agency, a hospital system, a university system — impose a time-limited moratorium on adopting a class of AI tool, pending completion of a defined review whose scope and outputs are set in the moratorium instrument itself. The vehicle is a district resolution, a chief judge's administrative order, a state agency directive, or a board policy that (a) suspends new adoption or expansion of the named AI class immediately, (b) commissions a specific review with a named body and a defined outputs list (harm assessment, community input, comparative safeguard analysis), and (c) sunsets on a defined date at which the moratorium either lapses, extends, or converts into a permanent policy. The strategy operationalises the precautionary principle at the public-sector layer where movement mobilisation can actually deliver an executive-level decision.

An actor chooses this strategy because it is politically easier to win than a permanent categorical ban — the public official who cannot say yes to "ban" can often say yes to "pause pending review" — and because the pause creates the procedural opening in which the deeper review, community-input processes, and safeguard-design work can happen without the deployment expanding underneath the movement's feet. The moratorium is also reversible: if the review finds the deployment safe under specific conditions the movement helps design, the moratorium lifts with those conditions in place; if the review finds the harms confirm the movement's concerns, the moratorium converts into permanent restriction with the review's findings as the record. And the sunset date gives the movement a scheduled contest point — the review's conclusion is a public moment the movement can organise around, converting an internal-agency decision into a public-facing accountability event.

It trades permanence for winnability. A precautionary moratorium is by design temporary and can lapse without the deeper policy change the movement sought, particularly if the reviewing body concludes the deployment is manageable with cosmetic safeguards. The strategy is also vulnerable to review capture — the body that conducts the assessment can be composed to reach the deployment-friendly conclusion, the terms of reference can exclude the movement's core concerns, and the output can be a report of ambiguous findings that the agency subsequently interprets in its own interest. Sustaining movement pressure across the moratorium's review period, and organising the community and expert inputs the review requires, is what determines whether the pause converts into a policy win or fades into a technocratic checkpoint the agency crosses.

Ecology

Distinct from municipal affirmative ban on a class of AI use. An affirmative ban is a categorical, permanent prohibition — no time limit, no sunset, no built-in review conclusion. A precautionary moratorium is time-limited by design and structured around a review that may or may not confirm the ban. Municipal facial-recognition bans (San Francisco, Oakland, Boston) are the affirmative-ban form; a hospital system's twelve-month pause on generative-AI patient communication pending safety review is the precautionary-moratorium form. The two are strategic-family variants — same instrument-target (public-sector deployment), different temporal-political shape.

Distinct from institutional procurement refusal of AI vendors. Procurement refusal is an ongoing procurement-policy stance: this institution does not buy from this vendor / does not buy this product class, indefinitely. A precautionary moratorium is a temporary hold on a specific deployment class regardless of vendor, framed around a review. The two operate on different institutional decision-surfaces — procurement offices' vendor lists vs. operational units' deployment plans — and typically involve different inside-institution interlocutors.

Feeds mandatory algorithmic impact assessment requirement. The review a precautionary moratorium commissions is essentially an ad-hoc AIA for the specific deployment. A moratorium regime that becomes routine practice — every new AI deployment automatically triggers a moratorium-and-review cycle — is a de facto AIA at the institutional level, and the codification of that practice into a standing requirement is the AIA strategy the movement pursues at statutory scale.

Distinct from mass protest for AI moratorium. Mass protest calls for an industry-wide pause on AI development at the frontier-lab level, on model training or deployment writ large. This strategy calls for institution-specific pauses on adoption at the public-sector deployment level. Both use "moratorium" language; they target different actors (labs vs. deployers) and different phases (development vs. adoption).