Microsoft's $20 Million Bounty Record Is Not a Like-for-Like Security Scorecard

That is a meaningful expansion of what the company is willing to pay for. It also narrows the inference readers can draw from the year-over-year dollar comparison: a program that accepts a wider set of security reports is not measuring precisely the same thing as the prior year’s program.

AI increased the reporting pipeline, not the proof of impact

Microsoft attributed a notable increase in submission volume in the second half of the year to researcher engagement and “the growing use of AI to support security research.” Its wording supports an account of AI as an aid to researchers; it does not say how many eligible reports involved AI, how many led to payouts, or whether AI changed the severity of the findings.

Zero Day Quest supplies a more concrete component of the record. The challenge and live hacking event brought researchers from 20 countries to Microsoft’s Redmond campus to work with security and engineering teams on cloud and AI scenarios. Participants submitted nearly 700 reports and received $2.3 million in awards, Microsoft said. That event accounts for an identifiable share of a total the company reports only as more than $20 million.

The result is a bigger discovery pipeline, but the supplied figures do not show its conversion rate: which submitted reports were reproducible, were in scope, produced fixes, or were duplicates. Those distinctions matter because bounty awards, report counts and resolved product risk are different measures.

A reward pool also depends on researcher confidence

The annual result arrives amid a public dispute that illustrates why total payouts are an incomplete test of a disclosure program. A researcher using the names Chaotic Eclipse and Nightmare Eclipse released details of several alleged zero-days without giving Microsoft time to patch them. The researcher alleged that Microsoft mishandled reports, ignored communications, withheld payments, deleted the reporting account and breached an earlier agreement; those are allegations, not findings established by the bounty data. The same independent report said some of the flaws were later exploited in the wild.

The reported conflict does not negate the program’s record or establish how typical one researcher’s experience is. It does show the distinct incentives at stake: coordinated disclosure gives a vendor an opportunity to assess and address a claimed vulnerability before public release, while public disclosure can shift that timetable.

Microsoft is not alone in using rewards to attract that research. A separate analysis says Google paid $17 million in vulnerability rewards in 2025, but that is a different company and reporting period, not a like-for-like benchmark for Microsoft’s July-to-June figure. The comparison is useful chiefly as a reminder that payout size is one element of competition for researchers, alongside scope and how reports are handled.

Microsoft’s company-reported number of researchers receiving bounty awards in the current and previous program years. Source: Microsoft Security Response Center.

What would show whether the bigger program works

Microsoft has provided aggregate payout, researcher and eligible-report counts, plus selected components of the record. Its announcement does not provide comparable measures for report triage, time to remediation, duplicate rates or the outcomes of AI-assisted submissions.

Those are the next decision-relevant figures. They would show whether the broader program is turning a larger flow of reports into addressed security risks—and whether the mechanisms intended to keep researchers reporting privately can scale with that flow. Until then, the $20 million record demonstrates higher spending and participation, not the speed or effectiveness of the full security response.