In early July, the House Budget Committee’s sweeping proposal, dubbed “One Big Beautiful Bill,” passed through the Senate and was signed into law by President Trump without a section in an earlier version that would have banned state and local governments from passing laws or regulations related to most artificial intelligence (AI) technology and automated decision-making systems.
Although the bill includes about $1 trillion in Medicaid cuts, excluding the AI regulation moratorium is a small win for the disability community. AI technology needs to be regulated to prevent bias against people with disabilities that can affect hiring processes, rent eligibility status, eligibility for government-run programs like Medicaid and Supplemental Nutrition Assistance Program (SNAP), and many other aspects of life.
Ariana Aboulafia, a project lead on disability rights and technology at the Center for Democracy & Technology (CDT), and Dr. Travis Hall, the State Director for the CDT, co-authored an article for Tech Policy Press on the potential harm the proposed AI regulation could do to the disability community before the moratorium, which was known as Section 43201 in certain drafts of the bill, was stricken.
Able News spoke to Aboulafia along with Henry Claypool, Technology Policy Consultant with the American Association of People with Disabilities, about why disabled people need to be aware of potential AI bias.
Aboulafia explained that AI and automated decision-making systems have a bias against people with disabilities because many of the systems were created with “underrepresented data sets” that excluded disabled people and other minority groups. Without data on the disability community to draw from, the technology will inevitably produce “under-inclusive outcomes.”
As an example, Aboulafia cited research from the University of Washington suggesting that resume screening and other AI technologies have been used in hiring processes for decades and have biases against resumes that include disability-related honors and awards. She said that in a hypothetical situation in which a resume screener is given two identical resumes, with the only difference being that one of the resumes has a disability-related honor listed, in some cases, technology excludes that resume, preventing it from being moved to the next stage, which usually includes a review by a human.
Aboulafia and Claypool note that technology used to screen for housing tenants has similar biases against the disability community. However, they are particularly concerned about how biases will impact eligibility for government assistance programs such as Medicare, Medicaid, and SNAP in the future.
“[Decision-making technology is] seductive because the AI tools can help you find people that are potentially eligible [for assistance], so you can enroll them in programs that will benefit [them]. But at the same time, if you turn them to make eligibility determinations and you’re not having the proper human oversight, the bias tends to come into play and give you discriminatory results,” said Claypool.
He noted that individuals denied Medicaid services due to AI-driven decisions have successfully sued their states, winning cases by proving that automated systems unfairly terminated their coverage. In a 2024 class-action lawsuit filed on behalf of 35 adults and children who were denied Medicaid coverage, a judge ruled that thousands of people in Tennessee had been wrongfully denied benefits, according to Gizmodo.
Claypool and Aboulafia explained that while there is no way to avoid AI and automated decision-making systems in today’s society, there are ways to diminish disability-related bias. For instance, companies like Enable Ventures are funding AI-related initiatives as part of the larger goal of closing the disability wealth gap.
Awareness of the disability bias in AI and the accessibility issues that prevent some disabled people from reaping the benefits of automated decision-making systems is the first step in making the technology fair, accurate, and inclusive.
Aboulafia told Able News she advises anyone using AI to make decisions in any sector to research the impact of this technology on all people, and marginalized people in particular, before using it.
“Anytime you’re going to use a tool, a good rule of thumb is to do both a pre- and post-deployment audit. So what that means is auditing the tool for things like racial bias, gender bias, disability related bias [before and after using it].”