The latest year changes the direction of the headline
Did VA’s reported harassment-related incidents keep rising in the most recent complete year? No. What’s the Scoop With Broach transcribed and summed the three annual series in figure 1 of a September 30 federal audit. Together, reported non-sexual harassment, sexual harassment and sexual assault fell from 7,129 in fiscal 2024 to 6,928 in fiscal 2025—a decrease of 201, or 2.8%. Each category declined, although sexual-harassment reports were almost unchanged.
The six-year comparison remains stark: the combined count rose from 2,591 in fiscal 2020 to 6,928 in fiscal 2025, an increase of 4,337, or 167.4%. Our original calculation separates the latest annual movement from the longer trend and identifies which categories produced each change. It shows why a rising-since-2020 description and a falling-in-2025 description can both be correct.
Neither establishes whether care became safer. The Disruptive Behavior Reporting System records incidents that people report or staff witness and document; it does not measure every occurrence. GAO found that VA had no systemwide effort to periodically establish the prevalence of these behaviors. More willingness to report, changes in actual behavior and other factors can affect these counts. The annual decline therefore cannot be credited to successful prevention from these records alone.
| Fiscal year | Non-sexual harassment | Sexual harassment | Sexual assault, alleged plus confirmed | Combined |
|---|---|---|---|---|
| 2020 | 583 | 1,701 | 307 | 2,591 |
| 2021 | 1,576 | 1,989 | 391 | 3,956 |
| 2022 | 1,915 | 2,519 | 470 | 4,904 |
| 2023 | 2,001 | 3,140 | 564 | 5,705 |
| 2024 | 2,415 | 3,969 | 745 | 7,129 |
| 2025 | 2,285 | 3,967 | 676 | 6,928 |
Almost all of the latest decrease came from two categories
Non-sexual harassment reports fell by 130, from 2,415 to 2,285; reports categorized as sexual assault fell by 69, from 745 to 676. Together those decreases account for 199 of the combined 201-report decline, or 99.0%. Sexual-harassment reports fell by just two, from 3,969 to 3,967. That is a 0.05% change, effectively flat at the precision normally used in public summaries.
The long-term comparison has a different composition. Sexual-harassment reports increased by 2,266 between fiscal 2020 and 2025, accounting for 52.2% of the combined increase. Non-sexual harassment added 1,702 reports and sexual assault added 369. Non-sexual harassment had the greatest percentage growth, approximately 292% using the displayed counts, but sexual harassment contributed the greatest number of additional reports. GAO describes the former increase as 291%; our calculation from its published endpoints is 291.9%, a difference that does not change the ordering.
Across all six years, the category totals reconcile to 31,213: 17,285 sexual-harassment reports, 10,775 non-sexual harassment reports and 3,153 sexual-assault reports. Those totals describe the database examined by GAO, which included incidents across 140 VA medical centers and associated outpatient clinics. They are not counts of unique victims, confirmed crimes or staff disciplinary findings.
A combined assault count must not become a count of proven assaults
GAO’s chart combines two sexual-assault statuses that the underlying VA system distinguishes. Of the 3,153 reports over the six years, 2,314 were categorized as alleged and 839 as confirmed or substantiated by VA Police. The report does not supply an annual alleged-versus-confirmed breakdown in that figure. The decline from 745 to 676 therefore cannot be described as a decline in confirmed assaults.
The distinction works differently for the two harassment categories. VA officials told GAO that behavioral threat assessment teams do not substantiate harassment or sexual-harassment reports because those teams do not conduct legal investigations. The categories reflect people’s reports of what they experienced or witnessed. Their function is to alert the teams so they can respond to safety concerns, including by recommending treatment for behavioral problems.
The system identifies patients generally. Most are veterans, VA officials told GAO, but some are non-veterans who received care, such as people experiencing a medical crisis on a VA campus. We retain that limitation when describing the audit’s scope. The publication has not obtained individual case files, determined guilt or identified people responsible for particular incidents.
Existing feedback tools do not resolve the measurement gap
VA already collects feedback through several channels. Its VSignals post-appointment customer survey asks about feelings of safety, and free-text responses can describe harassment. Officials told GAO they use artificial intelligence to identify sexually based concerns in those responses. But the safety question was not designed to measure harassment experiences, so its existence is not evidence of a representative prevalence survey.
The Harassment Veteran Feedback Project is more targeted. It began surveying women veterans in 2017 and added men in 2023. Between 21 and 35 participating medical centers have joined the survey each year. Participation is voluntary among facilities in a research network that currently has 77 members. VA researchers cautioned GAO that results are not generalizable and that changing participating facilities complicate trend comparisons.
GAO also notes that a proposed centralized platform could consolidate existing incident records. Combining databases could improve oversight of reported events, but it would still miss events that never reached an institutional reporting channel. That is a different problem from organizing the reports VA already possesses.
What the records support—and the unanswered accountability question
VA established its policy aimed at ending these behaviors in 2022. To assess progress, GAO recommended periodically identifying systemwide prevalence, potentially through representative surveys with questions describing specific behaviors. Such questions avoid relying solely on whether respondents label their experience harassment. Officials interviewed by GAO agreed that prevalence information was important for directing prevention and mitigation resources; VA supplied no comments on the draft report. The recommendation remained open on GAO’s public page when checked October 5.
Our examination is public-record analysis, not a survey of veterans or a new investigation of individual complaints. We transcribed every figure 1 value, reconciled the six annual sums and three category totals to 31,213, calculated endpoint changes and decomposed the latest decline. We then read the report’s definitions, survey limitations, agency-comment section and recommendation status. Fiscal 2025 was the latest complete year in the audit, not a measure of conditions in October 2026.
The comparison establishes that the latest decline was concentrated in non-sexual harassment and the combined assault category, while sexual-harassment reporting remained essentially flat. It leaves a concrete oversight question unresolved: did prevention reduce harmful behavior, or did the likelihood of reporting change? VA’s current evidence cannot distinguish those explanations systemwide. A favorable movement in a reporting total is therefore an observation requiring context, not proof that the policy achieved its goal.
Sources and further reading
GAO-26-109183, September 30, 2026: full report, figure 1 and methodology ↗
GAO current recommendation status, checked October 5, 2026 ↗