DREDF Is Mapping AI's Impact on Disabled Legal Clients

We build AI systems that make decisions about who gets housing, healthcare, and disability benefits. We have almost no systematic data on how those decisions are harming disabled people.

That gap is exactly what the Disability Rights Education and Defense Fund (DREDF) (opens in new window) is trying to close. The organization has launched a targeted survey of legal services providers — attorneys, paralegals, law students, intake staff — across eleven states, asking a deceptively simple question: are you seeing AI show up in your clients' cases?

The survey, supported by Borealis DIFxTech, targets legal aid organizations and law school clinics in California, Colorado, Connecticut, Florida, Idaho, Illinois, Mississippi, Nebraska, Ohio, Texas, and West Virginia. That's a deliberately broad geographic spread — rural states alongside urban ones, conservative jurisdictions alongside progressive ones — which suggests DREDF is after something more than a coastal snapshot.

What the Survey Actually Asks

The survey has two distinct aims, and both matter.

First: are legal services providers encountering cases where AI tools are affecting disabled clients' access to benefits, healthcare, employment, housing, or education? This is ground-level intelligence gathering. Legal aid attorneys are often the first people to see what algorithmic systems are actually doing to real people — not what vendors claim they do, not what regulators assume they do, but what happens when an automated eligibility determination lands on someone's doorstep and they can't afford to fight it alone.

Second: do those same attorneys understand AI well enough to recognize it when they see it, and do they want training? This is the more uncomfortable question. An attorney who doesn't know that an insurance company's prior authorization denial was generated by an algorithm can't challenge it as one. The survey is probing a knowledge gap that has direct consequences for disabled clients' ability to vindicate their rights.

Why This Moment, Why This Method

AI systems are already embedded in the decision-making infrastructure that governs disabled people's lives. Medicaid managed care organizations use predictive tools to allocate home care hours. Housing authorities use algorithmic screening to evaluate rental applications. Social Security's own systems use automated processing at various stages of claims review. The ADA (opens in new window) and Section 504 of the Rehabilitation Act (opens in new window) don't disappear because a decision was made by a model rather than a person — but enforcement becomes dramatically harder when the decision-maker is opaque.

The legal services community is positioned to see these harms first. But seeing them and being able to name them as AI-related problems are two different things. DREDF's survey is trying to establish whether that translation is happening — and where it's failing.

This kind of practitioner-level intelligence gathering is relatively rare. Most AI and disability research comes from academic institutions or policy think tanks working with aggregate data. Legal aid attorneys working in Mississippi or Nebraska have a different view: they see individual cases, individual clients, individual moments where a system failed someone who had no other recourse. Aggregating that ground-level knowledge is genuinely valuable.

The Knowledge Gap as a Civil Rights Problem

Here's the pattern worth examining: the communities most likely to be harmed by automated decision-making systems are often the least likely to have legal representation equipped to challenge those systems. Disability benefits claimants, Medicaid recipients, public housing applicants — these are populations that depend on legal aid precisely because they can't access private counsel. If the attorneys serving them don't recognize AI-driven decisions as a distinct category of legal problem, those decisions go unchallenged.

This connects to a broader dynamic that research on compliance implementation gaps has identified: legal victories and formal protections often fail to translate into actual change because the organizational capacity to implement and enforce them doesn't exist. The same logic applies here. Civil rights law may prohibit discriminatory algorithmic systems, but that prohibition is only as strong as the legal infrastructure available to enforce it.

The survey's secondary aim — assessing attorney interest in AI training — is therefore not a soft add-on. It's central to whether any of this matters. A practitioner who understands how to identify algorithmic decision-making, request documentation of the model used, and frame a challenge under existing civil rights frameworks is a fundamentally different advocate than one who sees only the denial letter.

What the Data Could Reveal

DREDF has stated that survey responses will inform an analysis of how disabled people are being affected by AI across multiple life domains, plus a report on legal services providers' training needs. That's a useful dual output — one document for policymakers and advocates, one for the legal services field itself.

The eleven-state scope creates an opportunity to surface geographic variation. Are AI-related harms concentrated in particular benefit systems? Do they look different in states with stronger disability rights enforcement infrastructure versus states with weaker ones? Do rural legal aid attorneys encounter different patterns than urban ones?

Those questions can't be answered yet. But they're the right questions, and this survey is one mechanism for beginning to answer them.

Key AI Decision Domains and Applicable Legal Frameworks

Life Domain	Common AI Application	Primary Legal Authority	Enforcement Pathway
Disability Benefits	Automated claim processing, fraud detection	Social Security Act; ADA Title II	Administrative appeal; federal court
Healthcare/Medicaid	Prior authorization algorithms, care hour allocation	Section 504; ADA Title II; Medicaid Act	State fair hearing; federal civil rights complaint
Housing	Algorithmic tenant screening, credit scoring	Fair Housing Act; ADA Title III	HUD complaint; federal lawsuit
Employment	Automated resume screening, performance monitoring	ADA Title I; EEOC guidance	EEOC charge; federal lawsuit
Education	Automated accommodations processing, proctoring software	Section 504; ADA Title II; IDEA	OCR complaint; due process hearing

For Legal Services Organizations: What to Do Now

If your organization operates in any of the eleven states named in the survey, participation is straightforward — contact bcho@dredf.org or call 510-644-2555, ext. 5234 to request the survey link. DREDF has noted that accommodations are available to complete the survey, which matters for staff with disabilities.

Beyond the survey itself, this is a good moment for legal aid leadership to ask some internal questions:

• Do your intake forms capture whether a client's adverse decision was made or influenced by an automated system?
• Do your attorneys know how to request algorithmic impact assessments or model documentation through discovery or public records requests?
• Has your organization identified AI-related harm as a distinct case category, or does it get folded into general benefits or housing dockets?

The DOJ's guidance on algorithmic discrimination (opens in new window) and the EEOC's technical assistance on AI in employment (opens in new window) provide starting frameworks, but the field is moving faster than formal guidance. DREDF's report, when it arrives, will likely be one of the more grounded practitioner-facing resources available.

The gap between what AI systems are doing to disabled people and what the legal system can currently see and challenge is real. This survey is one attempt to measure it. That's worth paying attention to.