Not My A.I.

Not My A.I. is the Latin American feminist-decolonial slogan-and-project-name through which civil-society research refuses Latin American public-service AI deployment in its current shape and reads it instead as the regional working surface of "oppressive A.I.". The framing was launched in 2021 by Coding Rights, the Rio de Janeiro-headquartered Brazilian feminist digital-rights organisation founded in 2015 by Joana Varon, in co-leadership with the independent Chile-based researcher Paz Peña. It operates simultaneously as a public-rallying slogan against the deployment of automated decision-making systems on Latin American welfare, health, and policing publics, as a project name with its own trilingual web home and mapping infrastructure at notmy.ai, and as the wider analytical register through which Coding Rights's Oppressive A.I.: Feminist Categories to Understand its Political Effects framework reads concrete Latin American AI deployments as questions of political power rather than of fairness or accuracy. The framing is the corpus's first Brazilian-anchored Message entry and, together with the adjacent msg-feminist-ai umbrella and the msg-data-colonialism interpretive frame, completes the cluster through which the regional Latin American feminist-and-decolonial AI-civil-society field carries its working public-discourse register.

Origin

The framing's working crystallisation runs through the 2021 launch of the Not My A.I. project as a feminist-toolkit and case-research line co-led by Joana Varon and Paz Peña under the institutional anchor of Coding Rights. The project's self-description, on its own home page, is that it is a "feminist toolkit to question A.I. systems" — a framing that does the slogan's substantive work in two registers at once: an analytical register, in which the project develops categories for reading AI deployment as a political artefact, and an organising register, in which "not my A.I." names the position from which Latin American publics, women, queer and trans people, and the poor refuse the deployment of automated systems built without their participation and used against their interests. The slogan's grammatical shape — a refusal in the first-person possessive — is itself the framing's substantive move: it names the AI deployments under examination as something done to particular publics rather than something done for them, and refuses the assumption that those publics are the ones whose consent and interests the deployments serve.

Coding Rights's own Why is Artificial Intelligence a Feminist Issue? programme page locates Not My A.I. inside the organisation's wider methodological frame ("Hacking the patriarchy since 2015") and supplies the substantive position from which the project operates: that "feminists' lenses go beyond the discussion of 'ethics', 'human-centered' or 'transparent' A.I. and enable us to check power imbalances", and that the working proposal is to "check power imbalances behind AI systems before deployment or development". Not My A.I. is the regional case-research vehicle through which Coding Rights operationalises that wider feminist-tech mandate against specific Latin American AI deployments, and the project's case archive is the corpus's clearest single record of the substantive case material on which the framing's analytical claims rest.

Oppressive A.I. — the analytical framework under the slogan

The analytical infrastructure underneath the Not My A.I. slogan is the Peña and Varon piece Oppressive A.I.: Feminist Categories to Understand its Political Effects, peer-reviewed and published through the Feminist AI Research Network's PubPub track. The piece's substantive contribution is to name three feminist analytical categories through which the political effects of AI deployment are read: oppressive intent (the deliberate design of AI systems to surveil, target, or punish particular publics — the most legible and the rarest of the three); oppressive effect (the impact AI systems have on particular publics regardless of intent — typically more substantial than intent-based analysis acknowledges); and oppressive design (the working features of AI systems' design, data, and deployment pipelines that route oppressive effects through ostensibly neutral or beneficial framings — the analytical category through which most Latin American public-service AI deployments are read in the project's case archive). The framework's working purpose, on the project page, is to "check power imbalances behind AI systems before deployment or development" — a deliberate move from the after-the-fact algorithmic-auditing register that anchors much US and European AI-accountability work to a before-the-fact refusal register that asks whether the system should be deployed at all.

The Oppressive A.I. register is the substantive analytical grain through which Coding Rights carries the wider msg-feminist-ai umbrella into the Latin American case-record. Where the Feminist AI umbrella names the cross-regional programme of work, and where the Afro-feminist AI register carried by Pollicy names the continental-African intersectional contribution, Oppressive A.I. is the Latin-American-feminist-and-decolonial analytical layer — anchored on case material from the region, articulated in the trilingual Portuguese-Spanish-English register of Latin American civil-society organising, and positioned as a deliberate refusal rather than a critique-from-within of the fair-and-inclusive-AI vocabulary.

The case archive — Salta, Projeto Horus, and Sistema Alerta Niñez

The framing's substantive working life is the case-study archive Not My A.I. has built across Argentina, Brazil, and Chile, with each case operationalising the Oppressive A.I. framework against a specific Latin American AI deployment. The framing's clearest single case artefact is Peña and Varon's 3 May 2021 case study on the Plataforma Tecnológica de Intervención Social — the Microsoft-backed adolescent-pregnancy-prediction AI deployed by the Argentine province of Salta from 2017 onwards and subsequently expanded to La Rioja, Chaco, and Tierra del Fuego, and ported to Brazil as Projeto Horus under a Microsoft proof-of-concept with the Brazilian Ministry of Citizenship. The case study reads the deployment in the verbatim register that this is "another expression of colonial extractivism" — Microsoft, in the case's framing, extracts data from Latin American social-protection programmes without documented accountability ("allegedly, not even the Brazilian government kept records of the results of the proof of concept, a closed box") and assumes "no responsibility for losses and damages of any kind" — and the case study's wider analytical claim names the class of Latin American public-service AI as designed to "automate, predict, identify, surveil, detect, target and punish the poor".

The case study's substantive critique runs across three interlocking layers. The statistical-and-methodological layer documents the Buenos Aires Laboratorio de Inteligencia Artificial Aplicada finding that the Salta system's results were "falsely oversized due to statistical errors" and that the database conflated reported pregnancies with actual prevalence — biasing the data toward populations with less access to safe abortion. The patriarchal-design layer reads the system's female-only data architecture as a working artefact of patriarchal design: the system "reinforces patriarchal gender roles and, ultimately, blames female teenagers for unwanted pregnancy" rather than naming the wider structural absence of sexual-and-reproductive-rights and education access the framing argues actually produces adolescent pregnancy. The automation-bias layer documents the working public-discourse register in which AI-mediated welfare decisions are received — that "if it is recommended by an algorithm, it is mathematics, so it must be true" — and the corresponding caseworker-level vulnerability to overvaluing algorithmic outputs above professional judgment. The three layers together do the framing's working analytical job: they read the deployment as a question of who builds the system, who its subjects are, what the system does, and on whose terms the public-record interprets it.

Beyond Salta and Projeto Horus, the Not My A.I. case archive carries Sistema Alerta Niñez — the Chilean Ministerio de Desarrollo Social predictive-risk algorithm for at-risk-children flagging — alongside several other ADM-on-the-poor case studies across the region. The case archive is the corpus's clearest single record of Latin American AI-on-the-poor deployment, and operates as the regional case-material counterpart to the empirical-survey record Derechos Digitales has built through its four-country comparative study on automated decision-making in Latin American public services (covering Chile, Colombia, Brazil, and Uruguay) under the wider msg-data-colonialism interpretive frame.

Relationship to the corpus's adjacent framings

Not My A.I. sits in a settled cluster with the corpus's two adjacent Latin American AI-civil-society Messages. The msg-feminist-ai umbrella names the cross-regional civil-society programme of work — the <A+> Alliance for Inclusive Algorithms, the Feminist AI Research Network, the regional academic hubs at Tecnológico de Monterrey, the American University in Cairo, and Chulalongkorn University, and the partner roster across Coding Rights, Pollicy, Derechos Digitales, and Digital Rights Foundation. msg-data-colonialism names the structural pattern — the Global-North / Global-South asymmetry of AI's extraction and accumulation — that the framing reads as a continuation of historical colonial appropriation. Not My A.I. is the working refusal that operationalises both: it carries the Feminist AI umbrella into specific Latin American case material, and it names the deployments the data-colonialism framing diagnoses as the things to which Latin American publics' answer is "not mine". The three framings travel together across the same regional civil-society infrastructure — Coding Rights and Derechos Digitales carry all three simultaneously, the Al Sur consortium of eleven Latin American and Caribbean civil-society organisations is the regional convening shape across which they circulate, and the Feminist AI Research Network's LAC hub at Tecnológico de Monterrey supplies the parallel applied-research register — and the corpus records them as the working analytical-and-rhetorical triad through which Latin American civil-society articulates its position on AI deployment.

Why it has carried

Three features have made the framing durable in the regional civil-society field over the 2021–2025 window.

First, the slogan's first-person possessive grammar does substantive work that the adjacent academic framings cannot. Where msg-data-colonialism names a structural pattern from an analytical position and where msg-feminist-ai names a programme of work from an institutional position, "not my A.I." names a refusal from a particular public's position — and the refusal's first-person possessive grammar is the rhetorical move through which Latin American publics' non-consent to the deployment is itself made part of the framing's substantive content. The slogan's grammatical shape encodes what the Oppressive A.I. framework names analytically: that the deployment of an AI system on a public who has not consented to it is itself a category of harm, regardless of the system's accuracy or fairness.

Second, the framing pairs slogan-and-analytical-framework with a working case-research archive in a way few framings of comparable scope manage. The trilingual Portuguese-Spanish-English notmy.ai infrastructure operates simultaneously as a public-rallying site, a case-study repository, and an open-mapping platform inviting civil-society contributors to register further regional AI-on-the-poor deployments — the same kind of distributed mapping infrastructure that the corpus's adjacent #BanTheScan framing operationalises in India through Project Panoptic and that Algorithm Watch carries in Europe. The case archive's working public life — across Salta, La Rioja, Chaco, Tierra del Fuego, Projeto Horus, Sistema Alerta Niñez, and the further Latin American deployments the mapping accumulates — gives the slogan its substantive analytical weight and prevents it from collapsing into a pure-rhetoric register.

Third, the framing's Latin American grassroots-feminist epistemic position refuses the assumption that AI-accountability vocabularies must travel from the Global North to the Global South. The Peña and Varon co-leadership — a Brazilian working in Rio de Janeiro and a Chilean working independently across the region — and Coding Rights's self-positioning as operating from "a collective, transfeminist, decolonial, and antiracist perspective of human rights" together carry the framing's working epistemological commitment: that the substantive contribution Latin American civil-society organising makes to global AI governance is being articulated from regional positions, in regional languages, on regional case material, in regional analytical traditions, and inside regional convening infrastructure — not as a translation of fair-and-inclusive-AI vocabularies into Spanish and Portuguese. The framing's Brazil-and-wider-Latin-America-anchored shape is, in this respect, what closes the corpus's previously-empty Brazil Message slot and supplies the regional-feminist-and-decolonial counterpart to the corpus's Anglosphere algorithmic-accountability and frontier-AI-safety framings.