Meenakshi is an engineering student in her final year, currently preparing for software job interviews. She is academically capable, resilient, and ready to enter the workforce. At 16, she survived a car accident that left her with a permanent locomotor disability. The accident also cost her a year of schooling.
She fought her way back. She resumed her studies. She completed her degree.
And now her job applications are being rejected out of hand—not by a person but by a system.
Today, most large recruiters rely on online application portals. Many of the top technology companies automatically filter out candidates who have “gaps” in their education. Their forms do not ask why such gaps exist. They do not allow for explanations. They do not take context into account. An algorithm simply marks her as ineligible.
How AI reproduces ableist bias
Research has documented that AI recruitment tools reflect ableist bias and treat educational gaps, nonstandard timelines for completion, differences in communication, and other disability-linked signals as indicative of poor candidate quality. A US lawsuit, Mobley v. Workday, alleges that a widely used AI hiring platform systematically screened out candidates with disabilities, and a federal court found that AI vendors, not just employers, may bear direct liability for such outcomes. In India, the stakes are particularly high: persons with disabilities have a workforce participation rate of around 36%, compared to 60% for those without. Automated hiring systems are widening, not narrowing this gap.
As recruiters increasingly outsource the initial vetting of candidates to AI-driven systems and third-party vendors, a clean academic timeline becomes a proxy for merit, while one with medical interruptions becomes a red flag.
Diversity in reports, exclusion in code
The irony of Meenakshi’s situation is difficult to miss. The same companies enthusiastically adopting AI models that discriminate against those with disabilities publicly champion diversity and inclusion. They publish sustainability reports. They affirm compliance with India’s Rights of Persons with Disabilities Act, 2016 and international commitments related to the United Nations Convention on the Rights of Persons with Disabilities (UNCRPD). They declare zero tolerance for discrimination.
Yet their recruitment systems silently exclude candidates like Meenakshi.
Her only realistic chance lies with smaller firms that still rely on humans capable of discretion in their review processes. The most technologically advanced organizations—the ones shaping the future of AI—become the least accessible.
An invisible form of discrimination
The most troubling aspect of AI-driven inequity is that no one believes a wrong has occurred.
Not college placement offices.
Not the companies looking for new hires.
Not their third-party, subcontracted vendors.
Not the regulators responsible for preventing discrimination.
Not even Meenakshi, who is busy proving herself as capable as any other candidate.
This is systemic exclusion—unhindered, automated, and largely unnoticed.
Reasonable accommodation in design
The principle of reasonable accommodation embedded in disability law in most countries requires employers to adjust processes to ensure equal opportunity. The UNCRPD, ratified by 186 countries, obligates states to prohibit discrimination in hiring and to ensure reasonable accommodation. The use of a neutral-seeming criterion—an unbroken academic record—that in practice disproportionately disadvantages a protected group—candidates with disabilities—may constitute indirect discrimination. The failure to build in an accommodation pathway is a legal failure, and employers who delegate screening to AI systems remain legally responsible for their outcomes.
But how does one seek accommodation from a drop-down menu? How does one explain disability to an application portal that has no field for explanation? Discrimination in the AI age does not require prejudice. It only needs poorly designed parameters.
Years may pass before researchers can document patterns of disproportionate exclusion. Audits may eventually reveal that candidates with disability-linked academic interruptions are underrepresented in interview pools. Committees will be formed. Policies will be revised.
Meanwhile, thousands of candidates will have lost the formative years of their careers.
The solution is simpler than the excuse
This is not a technologically complex problem. Recruitment systems can provide structured space to explain academic gaps, allow voluntary disability disclosure, and trigger human review instead of automatic rejection in such cases. They can conduct bias audits of AI shortlisting tools and align algorithmic filters with antidiscrimination law.
None of this requires new infrastructure. But it does require design choices that treat inclusion as a requirement, not an afterthought. Accommodation can and should be embedded at the design stage. If coding ability can be quantified, so can fairness metrics.
Why changes have not been made
If the solution is straightforward, why has it not been implemented?
Because the AI bandwagon is rolling ahead, while regulation lags. Because inclusion is treated as a branding exercise rather than a rights-based obligation. Because life, it seems, can go on without inclusion—but not without efficiency gains driven by AI.
Organizations wait for regulation.
Regulators wait for violations.
Victims wait for recognition.
The cycle sustains itself.
However, some jurisdictions are beginning to move. The EU’s AI Act, which entered into force in 2024, classifies recruitment screening tools as “high risk” and requires transparency, human oversight, and bias mitigation before deployment. The Institute for Human Rights and Business has documented how these tools entrench disability discrimination through design principles that assume all candidates follow identical life paths. Regulation and accountability are building. But the process takes time, and meanwhile harms accumulate quickly.
Toward human-centered AI
AI systems are not neutral tools. They are powerful decision-makers capable of reshaping access to employment, credit, education, and dignity. We must explicitly recognize their capacity to compromise human rights, particularly the rights of persons with disabilities. Employers have an obligation to assess discriminatory risk in any tool they deploy. Technology vendors cannot escape that obligation by describing their products as neutral platforms.
Human-centered AI cannot remain a slogan. It must be implemented in practice.
Otherwise, people like Meenakshi will continue to be treated not as talent, but as anomalous data.
She overcame a devastating accident.
She should not have to overcome a discriminatory algorithm.