The Government Accountability Office just handed the U.S. Patent and Trademark Office a very loud wake-up call. In its new GAO report (GAO‑27‑108491), the watchdog warns the USPTO’s push to shrink the backlog is likely to fail unless it stops losing patent examiners and rethinks a plan that trades training and quality for speed. The message is blunt: do more with less may sound efficient on a spreadsheet, but it is a bad plan for inventors, businesses, and the rule of law.
Backlog targets versus reality
The GAO found the backlog of unexamined patent applications swelled to roughly 838,000 and, even after some improvement, still sat near 756,000. The USPTO set a goal to get below 599,000, but GAO says that won’t happen before 2031 unless examiner departures fall. That matters. Long patent pendency means inventors wait months or years for clarity. Businesses can’t plan. Investors and manufacturers face extra risk. Promising a faster system while it’s understaffed and overworked is like advertising a high-speed train that keeps losing conductors.
Cutting training to chase metrics
Here’s the part you might find hard to believe: the USPTO cut annual subject-matter training and slashed new-examiner training in half, then told examiners to make up the time by processing more applications. Attrition rose from 2.9% to 7.6%. The agency hired about 1,054 examiners in a recent period but lost roughly 1,209 in the same stretch. Examiners are working unpaid overtime to hit new quotas. So management trims training, raises output targets, and then blames the workforce when people burn out and leave. That’s not leadership — it’s hubris with a dash of denial.
AI tools and the illusion of automation
The USPTO has rolled out AI search tools to speed work, but GAO reports those tools are often unreliable. Examiners told investigators that the Similarity Search tool sometimes returned incomplete or irrelevant results, forcing full manual searches anyway. At the same time, the agency has paused or cut back quality surveys. So the USPTO is leaning on immature AI, shrinking training, and removing quality checks — all while insisting the backlog will vanish. If that isn’t a recipe for lower patent quality, I don’t know what is.
What must happen next
Director John A. Squires and Congress need to act. Restore training hours, pause new productivity quotas until quality safeguards are back in place, fix AI tools before they become a crutch, and make the USPTO publish its quality metrics again. The GAO recommendation is straightforward: assess and address the risks to retention and examination quality. Inventors deserve a patent system that’s fast and fair — not one that sacrifices expertise for empty targets. If the agency wants results, it should stop mistaking motion for progress.

