Protected, But No Longer Measured

On that date, a federal regulation changed for covered federal contractors and subcontractors, the companies that hold contracts large enough to fall under Section 503 of the Rehabilitation Act. Nothing changes about who those companies are allowed to discriminate against. What changes is what the federal government requires them to ask, count, and compare.

Starting that day, a covered contractor no longer has to invite job applicants to say, voluntarily, whether they have a disability. It no longer has to collect data on how many applicants with disabilities applied for a job and how many were hired. It no longer has to compare the share of people with disabilities in each part of its workforce against the government's utilization goal. A rule finalized by the Department of Labor on August 21, 2026 removes all three requirements, along with the regulation that told a contractor how to interpret the results if it fell short.

This is a change worth understanding on its own terms, separate from the larger argument over the past two years about diversity, equity, and inclusion policy. It touches a real legal apparatus with a specific history, a specific court record, and a specific, carefully reasoned justification on both the side that built it in 2013 and the side dismantling it now. Neither side is straightforwardly right or wrong here in a way this article is equipped to settle, and it does not try to. What it can do is show, precisely, what the September 21 change does and does not do, and why the government that once defended this system in federal court now argues the opposite about the same law.

Since 2013, applicants at federal contractors covered by Section 503 have been offered a form known in government shorthand as CC-305. The form invited applicants, and later employees, to voluntarily identify whether they had a disability, for purposes of the Section 503 program. It appeared before a job offer, again after an offer was made, and periodically for existing employees. The government's own most recent estimate, approved by the Office of Management and Budget (OMB) in July 2026, put the form's annual volume at approximately 41.6 million responses a year across covered contractors nationwide, up from an earlier estimate of roughly 35 million, an increase the filing attributes to workforce growth and application volume rather than any change in who is covered.

Answering was voluntary. The form itself, required by federal regulation to say so in plain language, told the applicant that refusing to answer carried no consequence and that a response was not supposed to influence whether that specific person was qualified for the job. The information was meant to be kept on a separate, confidential record, apart from the file a hiring manager consults to make a decision about a candidate.

The purpose was never to screen any individual. It was to let a company add up the answers, at the level of an entire job group or an entire workforce, and get a number: what share of the people applying, and getting hired, identified as having a disability. That number fed into everything else Section 503 required.

When the Department of Labor's Office of Federal Contract Compliance Programs built this system in 2013, its argument was direct: a company cannot know whether its outreach to disabled job seekers is working if it never counts who is applying and who is getting hired. The rule introduced what the regulation called a utilization goal, set at 7 percent, and required contractors to compare their own disability representation against it, broken out by job group within the company rather than as one company-wide figure.

It is worth being precise about what this required and what it did not. If a contractor's disability representation in a job group fell below the goal, the regulation did not require the company to hire a specific person, adjust a test score, or fill a quota. It required the company to look for the reason: to review its recruitment sources, its application process, its outreach programs, and to document what it found. If it found a real barrier, it had to put a correction in place. The government's own enforcement data shows this obligation had some teeth: since 2020, OFCCP issued more than 75 violations specifically for a contractor's failure to do this analysis or follow through on the corrective step, not for missing the number itself. Missing the goal was a diagnostic trigger. Failing to diagnose and respond was the actual violation.

None of this establishes that the utilization goal prevented discrimination or proved it was happening at any particular company. What the record supports is narrower: the requirement obligated a contractor to collect data, look at it, compare it to a goal, document what it found, and respond. Each of those steps is documented in the regulation and in OFCCP's own compliance manual. Whether that obligation changed outcomes for disabled workers, as opposed to changing paperwork, is a separate question the available evidence does not answer.

A great deal of the argument around DEI policy over the past two years has turned on the word merit, usually as a stand-in for the idea that a hiring decision should rest on who is most qualified for the job. It is worth separating two things that get folded together in that argument: the standard used to judge who is qualified, and the infrastructure that determines who gets a fair chance to be judged by that standard in the first place.

The Section 503 utilization goal belonged to the second category. It never told a contractor whom to hire. It never adjusted a test score, waived a qualification, or created a preference for a less-qualified candidate over a more-qualified one. It measured whether disabled applicants were reaching the process at all, and whether the company's own outreach was functioning. A contractor that hired only the most qualified applicant in every case could satisfy this requirement in full, as long as it had looked honestly at whether disabled candidates were showing up to compete.

This distinction should not be overextended into a claim that everything labeled DEI worked the same way. It did not. Some mechanisms elsewhere in corporate and legal practice, unrelated to Section 503, did factor a protected characteristic directly into an evaluation or selection decision, which is a different thing and raises different legal questions under Title VII and related law. The utilization goal is not an example of that. Collapsing the two into a single undifferentiated category, in either direction, misdescribes what actually happened here.

"DEI" became a label wide enough to cover race and sex hiring preferences, disability accommodation, accessibility, outreach and recruitment, measurement and reporting, employee mentoring programs, supplier diversity initiatives, and internal affinity groups. These are legally and functionally distinct arrangements, built under different statutes, serving different purposes, carrying different risks. The retreat from DEI-labeled policy over 2025 and 2026 did not touch them uniformly, and a story that treats the whole label as one thing will get individual pieces of it wrong. This article is about one specific piece: the quantitative measurement architecture under Section 503.

The chain of events runs through several distinct decisions, not one.

Executive Order 14173, signed January 21, 2025, revoked Executive Order 11246, the decades-old order underlying the federal affirmative action framework for race and sex, and directed OFCCP to stop enforcing programs built on it. Section 503's own utilization-goal regulation, separately, contained an explicit requirement that contractors use "the same job groups established for utilization analyses under Executive Order 11246" when running their disability analysis. When the underlying order was revoked, that specific cross-reference had nothing left to point to. This is a real, documented dependency in the regulatory text itself, not simply two programs that happened to share paperwork.

Independently of that, the Department of Labor adopted a new legal position: that the mandatory invitation to self-identify a disability is itself a disability-related inquiry the Americans with Disabilities Act restricts. This argument does not rest on any change to the ADA's text. It rests on a 2024 Supreme Court decision, Loper Bright Enterprises v. Raimondo, which ended the deference courts previously gave federal agencies' reasonable readings of ambiguous statutes and required courts instead to determine the single best reading themselves. The department argues that under that stricter standard, its own earlier interpretation of the ADA does not hold up.

Separately again, a commenter in the 2026 rulemaking proposed a fix for the broken job-group cross-reference: write an independent job-group definition into the Section 503 regulation, rather than relying on the now-defunct Executive Order 11246 framework. The Department of Labor declined, citing the deregulatory directive of a different executive order, 14219, rather than any claim that such a fix was legally impossible. The job-group dependency was a genuine structural problem. The decision not to solve it was a separate policy choice.

It is worth noting what did not cause this change. Executive Order 14398, the order most associated in public discussion with the federal contractor DEI rollback, addresses race and ethnicity-based treatment specifically and is not the legal basis for the Section 503 changes described here. The disability-specific story runs through a different set of orders and a different legal argument entirely.

Asked for comment, a Department of Labor spokesperson said in a September 18 statement to Accibly Insights that the rule "does not diminish the commitment by the Department of Labor to empowering Americans with disabilities to be productive members of the workforce."

"What it does do is it streamlines several unnecessary requirements and data collections imposed upon federal contractors, saving over $80 million per year," the spokesperson said. That figure is a projection, not a measured result: the final rule's own regulatory analysis estimates annualized cost savings of approximately $80 million, in 2024 dollars at a 7 percent discount rate.

The spokesperson added that "OFCCP is still investigating complaints via the Section 503 process" and that the department continues to provide tools and resources to help employers "move beyond compliance to benefit from the value and talent disabled Americans bring to the workplace."

Beyond noting that complaint investigations continue, the statement did not directly address Accibly Insights' questions about how OFCCP will identify patterns of disability hiring discrimination without the removed data, or whether proactive Section 503 compliance evaluations will resume.

This is worth telling as a sequence, because the sequence is the story.

In 2013, when OFCCP first proposed requiring the disability self-identification invitation, employer groups raised exactly the objection the Department of Labor now embraces: that asking an applicant, before a job offer, whether they have a disability conflicts with the ADA's ban on pre-offer disability inquiries. OFCCP rejected the argument, calling it based on an incorrect reading of the law. It pointed to the Equal Employment Opportunity Commission's (EEOC) own interpretive guidance, which had specifically carved out federal contractor affirmative action self-identification from the ADA's inquiry restrictions, and to EEOC's technical assistance materials describing the same exception. To close off any remaining doubt, OFCCP obtained a letter from an attorney in EEOC's Office of Legal Counsel affirming that the practice was lawful.

Employer groups sued anyway. The case, Associated Builders and Contractors v. Shiu, reached the D.C. Circuit, which upheld the 2013 rule in 2014, under the deference standard courts applied to agency interpretations at the time. The Supreme Court declined to review the case in 2015.

In 2026, the Department of Labor takes the opposite position on the same statute. The ADA has not been amended in the interim. What changed is the legal environment surrounding how much weight a court gives to an agency's reading of a law it administers. The department's new rule states plainly that it no longer defers to any of the EEOC guidance it once relied on, including the 2013 letter, which it now describes as the non-binding opinion of a single career attorney, never put to a vote of the full Commission.

It is important to be precise about what has and has not been resolved. No court has yet ruled on the department's 2026 interpretation under the legal standard now in force. The 2014 decision addressed the earlier rule under a deference standard that no longer applies the same way. It settled the question that was in front of it. It did not settle the question the department is asking now, and nothing in the public record answers that question definitively in either direction.

The same day the Department of Labor finalized the Section 503 changes, it also finalized a rule addressing VEVRAA (the Vietnam Era Veterans' Readjustment Assistance Act), the parallel federal contractor law covering protected veteran status. The contrast between the two rules is one of the sharpest facts in this story, and it is worth stating carefully, because it is easy to describe in a way that distorts it.

This is not a comparison between veterans and disabled civilians. A disabled veteran can be, and often is, covered by the ADA, Section 503, and VEVRAA all at the same time. The comparison is between two parallel federal contractor accountability systems, issued by the same agency on the same day, that ended up in very different places.

VEVRAA's architecture survives intact. Contractors still must invite applicants to identify as a protected veteran before a job offer, and to identify more specifically, including as a disabled veteran, after an offer is made. They still must collect and retain data on veteran applicants and hires. They still must set an annual hiring benchmark and use that data, by regulation, as one of the criteria for judging whether their outreach is working. The rule the department issued for VEVRAA that day was a procedural cleanup: it relocated some administrative language and removed references to the now-defunct Executive Order 11246. It did not touch any of the substantive requirements above.

Section 503 lost the comparable architecture entirely: the self-identification invitation, the applicant and hire data collection, and the utilization goal.

The department's explanation for the divergence rests on two points that have nothing to do with which group the government values more. First, VEVRAA's hiring benchmark was never built on the Executive Order 11246 job-group structure the way Section 503's goal was, so it had nothing to lose when that order was revoked. Second, the department argues that asking someone whether they are a protected veteran is not, in itself, a disability-related inquiry, since most of VEVRAA's veteran categories have nothing to do with disability status. Where the form does ask specifically about disabled veteran status, the department argues that category, which turns on eligibility for a VA disability rating, is legally distinct from the ADA's own definition of disability, so it does not raise the same concern.

That argument is definitional. It does not attempt to estimate how much the two groups actually overlap in the real world, and the department's own rule offers no such estimate. A reasonable critic could point out that a sharp legal line drawn between two legally distinct categories may still leave a meaningful population caught on both sides of it.

Whatever the strength of the department's reasoning, one thing should not be misread from this comparison: VEVRAA's surviving data does not substitute for what Section 503 measured. It only reaches veterans, a subset of the broader disabled population, and even within that subset, its hiring benchmark tracks protected veteran hiring as a category. It produces no separate figure for disability specifically, even for veterans.

It is worth stating plainly, because the word gets used loosely in this debate: neither the old Section 503 utilization goal nor the surviving VEVRAA hiring benchmark ever functioned as a quota. Both were built, in their own regulatory text, as measurement tools meant to gauge whether a contractor's outreach and recruitment were reaching the intended population, not as hiring mandates with legal force of their own. The Department of Labor itself describes VEVRAA's benchmark as a way for a contractor to measure "progress toward achieving equal employment opportunity," language nearly identical to how the old Section 503 goal was framed. A contractor below either number faced a required conversation with its own data, not a legal obligation to hire a specific person.

None of the changes described above touch the legal rights that protect a disabled applicant or employee at a federal contractor. The Americans with Disabilities Act's nondiscrimination protections remain fully in force. Section 503's own nondiscrimination mandate remains in force. The duty to provide a reasonable accommodation remains in force, unchanged in substance. The requirement that a covered contractor maintain an affirmative action program at all remains, even though the quantitative core of that program is gone; a contractor still has to do outreach and review whether it is working in a qualitative sense. OFCCP retains full legal authority to schedule a compliance evaluation, to demand a contractor's records, including accommodation requests, applications, and interview notes, in whatever format it chooses, and to require specific numeric corrective benchmarks once it finds a violation. A contractor still cannot retaliate against someone for raising a concern. And any individual, whether or not they ever filled out a CC-305 form, retains the ability to file a complaint with OFCCP or a charge with the EEOC, which can result in relief without that person having had to prove a pattern across an entire company.

A disabled applicant does not lose federal legal protection on September 21, 2026. That claim would not be supported by anything in this record.

The practical consequence is narrower than it might first appear, and worth stating precisely. The federal requirement that produced standardized, contractor-level disability data disappears, and the Department of Labor has not created any replacement measurement system to take its place. That does not mean disability employment becomes unknowable everywhere. National population-level data from the Bureau of Labor Statistics and the Census Bureau continues, unaffected. Private organizations, individual employers, and researchers remain free to collect their own disability employment data voluntarily, and some already do. What disappears is the mandatory federal channel that once produced a standardized, contractor-by-contractor picture, and nothing has replaced it.

The strongest challenge to the story above deserves real space, because it complicates any simple reading of what just happened.

The old system's own demonstrated results were modest. Across fiscal years 2020 through 2024, OFCCP's compliance evaluations produced monetary relief for at least 29 and no more than 45 individuals on a disability basis (a confirmed 25 in fiscal year 2021 alone), out of 124,096 people who received class relief on any basis across that same period. Whatever the utilization-goal apparatus was doing, it was not converting into disability discrimination findings or monetary recoveries at any meaningful scale.

A second complication cuts the same direction. By OFCCP's own public account, the agency had already stopped scheduling new Section 503 compliance reviews and closed every review that was already underway, starting in 2025, more than a year before this rule takes legal effect. That is a description of the agency's current administrative activity, not a statement about its underlying legal authority, which is unchanged and could support a resumption of scheduling at any time. In that narrower sense, the September 21 change formalizes a stoppage in practical enforcement activity that had, functionally, already happened.

Put together, a reader could reasonably conclude that many disabled applicants may not notice any practical difference because of this specific regulatory change. That conclusion should not be dismissed as undermining the story. It changes what kind of story this is. This is not, on the evidence available, a story about active enforcement collapsing overnight. It is a story about a government choosing to make itself, and the contractors it regulates, less able to see something they had previously been required to look at, in a period when the practical machinery for looking had already gone mostly quiet.

Disability hiring discrimination is not a new or contested finding. A 2018 field experiment sending more than 6,000 job applications found that disclosing a disability reduced employer interest by 26 percent. A broader academic literature, some of it dating to the 1970s, has documented disability hiring discrimination consistently across decades and countries. Separately, EEOC data shows disability charges reached the highest level on record in fiscal year 2025, a trend that began rising in 2022, three years before any of the policy changes described in this article.

None of this should be read as evidence that the current policy changes caused disability discrimination, or that rising charge numbers are a response to these rules. The timeline forecloses that reading entirely. The reason this evidence matters here is narrower and more precise: the question this article is asking is not whether disability bias exists. It does, and it predates this story by a wide margin. The question is what happens to the ability to catch it, in a labor market where it was already documented, once one of the channels that could surface a pattern without waiting for an individual complaint goes away.

A reasonable person following this story might ask whether reduced proactive accountability creates more room for arbitrary bias against disabled applicants. That is a fair question, and it deserves a precise answer rather than either dismissal or alarm.

The honest answer, based on everything examined here, is that this is a plausible risk mechanism, not an established outcome. Removing a numerical trigger for self-examination is the kind of change that, in principle, could give an employer more discretion, and discretion is where documented bias tends to operate with the least friction. But no one has measured whether employer behavior actually changes once this specific accountability disappears, because the rule affecting it has not yet taken effect. This is not a case where the policy grants anyone permission to discriminate. It grants no such thing, and nothing in Section 503, VEVRAA, or the ADA changes that. What would need to be true to answer this question with real evidence by 2027 or 2028: a measurable shift in disability-related EEOC charge patterns beyond the trend that already existed; data on whether contractors continue collecting disability information voluntarily; and evidence, one way or another, about whether outreach or accommodation investment changes. None of that evidence exists yet.

A right is something a person can bring to a lawyer or a federal agency after something has already gone wrong. A guardrail is something built to notice a pattern before any individual has to come forward and prove it. The disability employment story unfolding through September 2026 is, at bottom, a story about a right that survived and a guardrail that narrowed.

That framing should not be mistaken for a claim that the guardrail worked well, or that its disappearance leaves nothing standing. The guardrail's own track record was modest. The broader systemic authority behind it, the power to schedule a review, demand records, and impose a numeric fix once wrongdoing is found, was not removed. What disappeared is not the government's power to act on disability discrimination at a federal contractor. It is one of the ways anyone, including the government itself, might have found out there was something worth acting on in the first place.

The federal regulatory story above sits inside a much broader corporate retreat from DEI-labeled programs over the past several years, and the two should not be conflated. Walmart made its major changes to its diversity programs before the 2025 executive actions took effect, which means they cannot be described as compliance with orders that did not yet exist, evidence that the corporate movement had causes independent of federal policy. At the same time, other companies retained substantial inclusion programs under the same public pressure: both Apple and Costco kept their programs in place after their own shareholders voted, by margins above 97 percent, against proposals to eliminate them, showing that meaningful corporate discretion survived.

Among the three companies discussed above, none was found to have explicitly ended a disability or accessibility program during this period. This is a narrow observation about Walmart, Apple, and Costco specifically, not a claim about a broader sample or about every American employer, and it does not resolve whether broader cuts to general inclusion programs could still affect disabled workers indirectly. It does mean the evidence available here does not support treating the corporate retreat, or the treatment of disability-specific programs within it, as one coordinated, federally caused event.

ADA nondiscrimination: in force before September 21, 2026, and in force, unchanged, after.

Section 503 nondiscrimination: in force before, and in force, unchanged, after.

Reasonable accommodation duty: in force before, and in force, unchanged, after.

Section 503 affirmative action program: required, with a quantitative core, before; required, with that quantitative core removed, after.

Disability self-identification invitation: required before; no longer required after.

Applicant and hire disability data collection: required before; no longer required after.

Utilization goal analysis at 7 percent: required before; removed after.

Outreach effectiveness review: required to weigh quantitative data before; that data-weighing requirement removed after.

OFCCP records and evaluation authority: in force before, and in force, unchanged, after.

Individual complaint and charge rights: available before, and available, unchanged, after.

Effect of the 2026 rule on employment outcomes: not applicable before, since the rule was not yet effective; unknown after, until post-effective-date evidence develops.

Documented change: the quantitative measurement architecture described above is being removed by regulation, on a fixed date, and no substitute federal data source exists to replace it.

Plausible risk, not established outcome: reduced quantitative accountability increases employer discretion in a labor market where disability bias is already documented; some outreach incentive is weakened by the removal of the data-weighing requirement; oversight becomes more dependent, in practice, on individuals recognizing and reporting discrimination themselves, though the legal authority to act without a complaint remains in place.

Unknown: whether actual hiring or retention outcomes for disabled workers change at all. No data capable of answering that question exists yet, because the rule has not taken effect. Predicting an answer here would outrun the evidence.

A short, concrete list of indicators that would eventually let someone answer the question this article cannot answer now: whether OFCCP resumes scheduling Section 503 compliance evaluations, a program that has been dormant since 2025; EEOC's disability charge data in the years after this rule takes effect, watched for any deviation from the trend that predates it; whether contractors continue collecting disability self-identification data voluntarily, despite no longer being required to; participation and self-identification trends reported by Disability:IN, the closest thing to an independent private benchmark; whether any pending litigation over the broader executive order framework changes the legal landscape; and the Bureau of Labor Statistics' ongoing disability employment and labor force participation data, which will continue regardless of what happens at the contractor level.

Most of this story belongs to a narrow, specific regulatory change, and it should not be forced onto every corporate example or broader DEI development this investigation touched. But at its center is a genuinely useful question, one that applies cleanly to Section 503 and nowhere else in this reporting quite so precisely: what happens when an obligation survives, but the instrument that was used to measure whether it is actually working disappears.

After September 21, 2026, the law still says a qualified person cannot be denied a job because of a disability. That has not changed and this article has found nothing to suggest it will. What changes is how much a federal contractor is required to know, and required to show anyone else, about whether disabled people are actually reaching its workforce in the first place.

The right was never really the question here. The question is who will notice if the right, on its own, turns out not to be enough.

Accibly Insights contacted the Department of Labor's Office of Public Affairs on September 14, 2026, with questions about the Section 503 rule and requested a response by September 25.

A department spokesperson provided the statement quoted above on September 18.

Accibly Insights sent follow-up questions on September 20 about how OFCCP and federal contractors will assess whether affirmative action efforts are effective without the former utilization and disability data requirements, and whether OFCCP will publish aggregate information on disability employment outcomes among covered contractors.

No response to those follow-up questions had been received as of October 13, 2026.

Accibly Insights separately contacted Disability:IN on September 14, 2026, requesting written comment on what removing this measurement system changes in practice, with a requested response date of September 25.

No response was received by the deadline.

Recommended Actions

• ✓Track OFCCP's Section 503 scheduling activity for any resumption.
• ✓Track EEOC's disability charge data past FY2026 for deviation from the pre-existing trend.
• ✓Monitor Disability:IN's future reporting for any change in voluntary self-identification participation.
• ✓Watch pending litigation over the broader executive order framework for any spillover effect on this specific rule.