Nothing About Us Without Us

"Nothing About Us Without Us" is a disability rights principle that has become a foundational democratic governance demand in AI accountability contexts: no AI system, policy, or governance framework affecting a community is legitimate unless the communities it most affects were meaningful participants in its design, deployment, and evaluation. The phrase does not argue that AI systems should be designed for affected communities — it demands they be designed with and by those communities, with participation structured as a governance requirement rather than a consultation add-on. In the corpus it sits alongside Indigenous data sovereignty, feminist AI, and Consent, Credit, Compensation as a community-authority demand distinct from the accountability and harm-documentation framings that dominate algorithmic justice discourse; where most algorithmic accountability work asks AI systems to do less harm, "Nothing About Us Without Us" asks who gets to decide.

Origins in disability rights

The phrase carries multiple historical layers. Its deepest documented root is the Polish Nihil novi constitutional principle of 1505, establishing parliamentary authority over monarchical decree — a democratic-participation tradition whose force the disability rights movement would eventually press into service across a very different institutional context. Its modern form was transmitted through South African activists: James Charlton first heard the phrase from Michael Masutha and William Rowland — leaders of Disabled People South Africa — who had themselves heard it from an unnamed Eastern European activist at an earlier international disability-rights conference. Charlton's 1998 book Nothing About Us Without Us: Disability Oppression and Empowerment (University of California Press) codified the principle as the organising demand of a global disability rights movement: that disabled people hold the only authoritative voice on the policies and systems affecting their lives, and that governance conducted without that participation is not merely inadequate but oppressive. David Werner published a book with the same title in the same year. The United Nations adopted the phrase as the theme for the International Day of Persons with Disabilities in 2004, extending its institutional reach into multilateral governance contexts.

The disability justice framework, developed in the early 2000s by the collective Sins Invalid and grounded in principles including leadership of the most impacted, intersectionality, and collective access, extended the participatory demand to the intersection of race, class, gender, and disability. The most marginalised within any affected community must lead — not merely be consulted — on governance decisions affecting their lives. This intersectional layer is the frame through which disability-justice advocates carried "Nothing About Us Without Us" into AI governance: the communities experiencing compound AI harms, including disabled people of colour, workers with non-apparent disabilities, and people reliant on automated welfare systems, must govern those systems, not merely survive them.

Arrival in AI governance

The phrase's entry into AI-specific advocacy tracked the arrival of AI systems in the domains where disabled people most directly face institutional authority: welfare benefits, employment screening, and criminal risk assessment.

Algorithmic welfare and benefits (2018–2022)

The first sustained AI governance application of the demand concerned algorithmic benefits-determination systems. The Center for Democracy and Technology's Disability Rights in Technology Policy programme, led by Policy Counsel Ariana Aboulafia, documented in October 2020 how algorithmic systems were cutting Medicaid-funded home-care hours for disabled individuals in Arkansas and elsewhere — automated determinations that reduced care without adequate human review or contestation pathways. The CDT's framing applied the NAUWU principle as a legal argument, contending that algorithmic non-participation violated ADA and Section 504 protections requiring that disabled people have meaningful access to the benefits they are entitled to claim and contest. The Allegheny County child-welfare AI system, subsequently under Department of Justice investigation, added a parallel case: an algorithm used to inform family-separation decisions had been built without participation from the communities — predominantly low-income Black families — it most affected.

Disability data exclusion and AI development (2022–2024)

Research in 2023 documented that every major NLP model tested exhibited significant bias against disability. Maitreya Shah (Berkman Klein Center for Internet and Society) found that disability was being systematically treated as "outlier data" excluded from AI training sets, and put the governance failure directly: "Do people with autism or other disabilities even want these technologies? No one asks them." Lawrence Weru (Harvard Medical School Biomedical Informatics) documented voice-recognition failures for users who stutter, showing that commercial AI assistants could not process his speech when he stuttered — making critical assistive-technology functions inaccessible by design. Naomi Saphra (Kempner Institute, Harvard) required customised voice-to-text software to use AI coding tools at all, having found commercial tools optimised for non-disabled interaction patterns. Each case illustrated the structural failure the NAUWU demand names: systems built without participation from the communities they are supposed to serve.

The Mozilla formulation (2024)

The most direct deployment of the phrase in AI governance advocacy came in Kenrya Rankin's July 2024 Mozilla Foundation blog post explicitly titled "Nothing About Us, Without Us: Disability Justice and AI." Rankin, writing for the Disability Culture Lab, named the governing demand as placing "the folks most impacted by AI in charge of regulating and visioning it" through organising — framing the demand not as a request for consultation but as a power question requiring political organisation, since government and tech companies will not voluntarily surrender power. The post named ableism embedded in AI training data as the mechanism through which exclusion perpetuates itself: systems trained without disability representation encode disability as deviation, and that encoding propagates into the downstream systems — healthcare algorithms, hiring tools, content moderation — that make consequential decisions about disabled people's lives. Rankin's formulation made explicit the connection between the disability rights principle and the broader AI governance demand: "If we are at the table when AI systems are created and deployed, we can help account for the needs of all" — the most-affected-community standard as the design criterion, not an accommodation add-on.

Why it carries

Three features explain the phrase's traction in AI governance specifically.

First, the demand is structural rather than instrumental. Most AI accountability work asks systems to perform better on fairness metrics — reduce disparate error rates, remove discriminatory proxies, audit for bias. "Nothing About Us Without Us" asks a prior question: who decides what performance means, which metrics are measured, and whether the system is built at all? The structural demand is harder to satisfy with a benchmark improvement. It requires redesigning the governance architecture of AI development to include community participation in the decision loop itself, which creates friction with the speed and commercial pressure under which AI systems are currently built.

Second, the disability rights origin grounds the demand in established legal rights rather than novel claims. The ADA, Section 504, and the UN Convention on the Rights of Persons with Disabilities (CRPD) create participation entitlements; applying them to AI governance converts the participatory demand from an ethical preference into a legal argument that regulatory enforcement mechanisms can act on. The CDT's framing of algorithmic benefits determinations as ADA violations illustrates the litigation pathway: the NAUWU principle, when grounded in disability law, gives AI governance advocates a route that purely technology-accountability framings do not have.

Third, the phrase's breadth — "nothing about us" rather than "nothing about disabled people" — has made it available to AI governance advocates across multiple affected communities. The related Indigenous data sovereignty framing parallels the demand from a sovereignty register; feminist AI carries the centering-affected-communities demand from a gender-justice tradition; Consent, Credit, Compensation formalises participation as an economic and IP claim. Together they constitute a cluster of community-authority demands whose structural target is the same — who governs AI — even as their legal, political, and movement grounds differ.