Coin and propagate a frame — a short phrase, a coined term, a slogan, a metaphor — that reorders how the press, the public, and policymakers see an AI system. "Stochastic parrots", "the coded gaze", "algorithmic colonialism", "no tech for ICE", "ban biometric mass surveillance": each names something that was previously diffuse and gives the movement a unit of shared vocabulary that travels through speeches, headlines, hashtags, and citations. The strategy is the deliberate, repeated authoring of these units, and the institutional support — papers, exhibitions, talks, op-eds — that anchors them.
An actor chooses framing because every other strategy in the movement depends on it. A frame that lands changes what counts as the default reading of an AI deployment, which in turn changes what regulators feel licensed to do, what journalists feel licensed to print, and what a target organisation can credibly defend. Insiders without a frame lose the rhetorical ground that organisers and researchers spent years preparing.
It trades off durability for reach. A frame can be co-opted (industry adopts "AI safety" and recasts it), inverted (opponents pick up the language with new meaning), or saturated (a slogan loses bite from over-use). Frames also do not, by themselves, change material conditions — they are necessary, not sufficient, infrastructure for the other strategies.
Verdict — wins discourse, exposed to capture
Counter-narrative framing is the corpus's most structurally upstream strategy. Every other strategy in the movement is delivered through the public-affairs register a frame establishes, and the strategy's adopter set runs the widest cross-section of any in the corpus: the Algorithmic Justice League (the coded gaze; your face is not your password), the DAIR Institute (stochastic parrots; data colonialism), Coding Rights (not my AI; feminist AI), MediaJustice (carceral tech; automating banishment), European Digital Rights (ban biometric mass surveillance; Reclaim Your Face), Stop Killer Robots (killer robots; meaningful human control), Mijente (no tech for ICE; take back tech), Fight for the Future (Ban the Scan; bossware; ban surveillance schools), PauseAI (pause giant AI experiments), and Amnesty International (Ban the Scan as a global frame; biometric mass surveillance). The four effects: above are typical of what the strategy returns to the movement — a multilateral UN resolution adopted in the coalition's own vocabulary, a critical frame that became the operative public-affairs register for the LLM moment, a binding EU regulation whose prohibition language was coined by the coalition lobbying for it, and the mass-public mobilisation that the post-GPT-4 frontier-AI moment registered as.
The strategy is good, with a failure mode that comes from working.
Strong on its primary claim — that a frame coined in the movement's vocabulary can become the default vocabulary the policy field, the press, and the public use to think about a class of AI deployment. The Stop Killer Robots framing carried the autonomous-weapons file from a specialist humanitarian-disarmament concern in 2012 to a 152-state-supported UN General Assembly resolution in 2023 without rewriting its terms. The Reclaim Your Face / Ban Biometric Mass Surveillance framing supplied the substantive prohibition language the EU AI Act ratified, and the European Citizens' Initiative the framing organised was the public-affairs ground the trilogue negotiators were standing on at the closing stage. The "stochastic parrots" frame redefined what a large language model was conceptually doing for the post-2020 LLM moment, and the Bender et al. paper has accumulated the highest in-field critical-citation count of any single artefact in the post-2020 AI-ethics literature. The "Pause Giant AI Experiments" framing produced the post-GPT-4 mass mobilisation cycle and seeded a federated protest movement that has carried the frame into Paris, London, San Francisco, and Brussels through 2025. The "No Tech for ICE" framing produced the documented onward lineage of "No Tech for Apartheid" worker organising at Google and Amazon over Project Nimbus in 2022–2024, the framing's adopters explicitly modelling their name on the original.
Weaker on its secondary claim — that discourse wins convert into material wins on the strategy's own. The Stop Killer Robots framing is thirteen years old and the multilateral system has produced a request-for-views resolution, not a negotiating mandate. The Pause framing did not pause anything: every named frontier-model provider has continued training successor models without an externally enforced break. The Ban Biometric Mass Surveillance framing produced a partial rulebook with the law-enforcement, migration, and national-security carve-outs precisely where the abuses are worst — the coalition's own public assessment of the EU AI Act has been that the law "took steps to legalise live public facial recognition across the bloc". The "stochastic parrots" critique has not slowed industry-scale LLM training one quarter; it produced a discourse the industry now navigates by publishing its own safety cards while continuing to scale. The strategy generates the working public-affairs register the downstream strategies use; whether the downstream arms can convert that register into a binding constraint is the open question this strategy cannot answer alone.
The deepest structural problem is the frame-laundering loop. The strategy's success rate at coining vocabulary outpaces its rate of retaining control of that vocabulary, and the framing target — the AI industry, at communications budgets in the high hundreds of millions of dollars annually — can adopt a critical frame at industrial scale and use it to defang the very regulation the original frame was supposed to enable. "AI safety", "Responsible AI", "Trustworthy AI", "Frontier AI safety", "AI alignment", "Model evaluation", "AI governance": every one of these was coined inside the movement or at its research-adjacent edges, and every one of these is now the operating vocabulary of a major frontier-model provider's product-trust, public-policy, or safety-team self-description. The Stochastic Parrots frame is the limit case — Google's response to a critical frame was to remove the authors, then build a Google-internal "Responsible AI" team that adopted the same surface vocabulary while abandoning the critical content. The strategy's structural exposure is not platform opposition but platform fluency: the frames the movement coined are the frames the industry now uses to talk about itself, and the framing arm of the movement has not yet found a stable answer to the question of what a critical frame is supposed to do once the target is publishing one too.
The strategy is bad as the lone arm. A coined frame without an institution behind it dies in saturation within three-to-five years; a frame with an institution but without empirical audit numbers anchoring it is a slogan; a frame with audit numbers but without coalition lobbying carrying it into a legislative process is a citation; a frame in a legislative process without strategic litigation enforcing the law's narrow remedies after passage is a paper rulebook. The framing arm is the upstream evidentiary register; the rest of the movement is the apparatus that converts the register into a constraint, and a frame deployed alone produces a rhetorical victory the industry can absorb at the cost of one product-trust team and one set of model-card releases.
Ecology
The strategy sits structurally upstream of every other in the corpus and feeds them in turn. It is fed by empirical audit and expose at the substrate: the audit number is what gives a frame attribution-grade citability — the Gender Shades 34.7 per cent error rate is what gives "the coded gaze" its political weight, the OpenSCHUFA findings give "black box" its anchoring, the Worker Info Exchange Managed by Bots findings give "algorithmic management" its evidentiary base. An audit without a frame is a table; a frame without an audit is a slogan; the audit-frame pair is what travels.
It is paired with survivor-led testimony as evidence as the complementary face-to-the-frame register: the audit produces the number, the frame names what the number means, the survivor supplies the face the number and the frame both attach to. Robert Williams of Detroit, Daniel Motaung in Nairobi, the Mexican human-rights defenders whose phones the Citizen Lab forensically attributed to Pegasus — each is a named survivor whose story carries the frame from a statistical claim to a moral one in the public record.
It uses open-letter collective signatory action as its principal launch vehicle. Many of the corpus's most consequential frames entered the public record as open letters: "Pause Giant AI Experiments" was the title of the Future of Life Institute letter that produced the frame; the Hollywood Artists 4 Safe AI framing entered the entertainment-industry record through joint letters from SAG-AFTRA and creative-industry coalitions; the EDRi-coordinated November 2021 fundamental-rights statement that anchored the Ban Biometric Mass Surveillance framing in the EU AI Act process was an open letter at form. The open-letter vehicle is what gets a frame into the press cycle at launch with the coalition's substantive vocabulary intact.
It feeds coalition lobbying of binding regional regulation as the public-facing register the lobbying coalition operates in. The EDRi-coordinated EU AI Act work ran on the "Reclaim Our Public Space. Ban Biometric Mass Surveillance!" framing; the Encode Justice / California State Senate work on SB 1047 and SB 53 ran on a frontier-AI-safety framing inherited from FLI; the ControlAI UK parliamentary campaign runs on an existential-risk framing it co-deploys with the protest arm. A lobbying coalition without a public-facing frame has no mass-political-economy register to point legislators to.
It feeds strategic litigation against algorithmic state decisions as the discourse register in which a deployed system's harm is named: the "coded gaze" framing carried into the ACLU's Williams v. Detroit Police Department wrongful-arrest filing as the way the disparate impact was characterised in the public-affairs cycle around the suit; the "algorithmic management" framing the Worker Info Exchange audit carried became the working register of the Amsterdam Court of Appeal's 4 April 2023 Uber and Ola rulings; the Pegasus framing the Citizen Lab attribution work coined is the working register of the multi-jurisdictional Pegasus civil and criminal filings.
It is merged with humanitarian disarmament treaty in the Stop Killer Robots case, where the public-facing frame is the coalition's institutional name and the substantive policy demand at the same time — the coalition's slogan and its UN-correspondence vocabulary are the same words. It is merged with mass protest for AI moratorium in the PauseAI / "Pause Giant AI Experiments" case, where the protest movement's name and the frame's verb are the same word, and the strategy's adopter is identical to its framing institution. These mergers are the strategy's highest-coherence form — frame, organisation, and demand fused into a single artefact — and produce the strategy's most concentrated public-affairs presence at the cost of the strategy's flexibility to switch frames when one degrades.
It feeds creator-class collective bargaining on generative AI through the "existential threat to creative professions" framing that anchored the 2023 SAG-AFTRA strike call and through the "Have I Been Trained?" framing the creative-industry generative-AI dispute organised on. It feeds map the AI supply chain of state harm through the "No Tech for ICE" framing that Mijente operationalised at the same time as the underlying supply-chain-mapping report — frame and map were launched together and the framing carried the map into the public-affairs register.
The strongest competing strategy is the AI industry's counter-framing-at-industrial-scale — the Responsible AI / AI Safety / Trustworthy AI vocabulary the labs publish and reproduce at communications budgets the movement cannot match. This is not a separate strategy with a separate adopter set; it is the same strategy run by the framing target against the framing organiser, and the movement's structural disadvantage in the contest is the budget asymmetry. The strategy's strongest hedge against the counter-frame is not louder framing but the downstream arms — an audit number, a court ruling, a binding regulation — that hold the original critical content in place when the surface vocabulary has been adopted by the target. The framing arm and the material arms are not substitutes; they are the same evidentiary chain held against the same target by different ends.