CybersecurityVulnerabilities & Patches
Microsoft security flaws strain under AI bug flood
Microsoft has fixed an estimated 5,383 security flaws in 2026 so far, up from 1,180 for all of 2025, according to GeekWire. The surge, driven by AI-assisted bug reports, is testing the company's patch process and raising stability concerns.

Key takeaways
- Microsoft patched an estimated 5,383 security flaws in 2026, up from 1,180 in 2025.
- The surge began around April and is driven by AI-assisted bug reports from new researchers.
- Average bug severity has fallen, but the volume is overwhelming a human-led patch process.
- Microsoft leaders have warned that rapid patching can destabilise older systems.
In this article
What happened
Microsoft security flaws are being patched at a record rate, with GeekWire reporting an estimated 5,383 fixed in 2026 so far. That is far more than the 1,180 flaws patched in all of 2025. The increase began around April, when security researchers started sending far more bug reports to the company.
How Microsoft handles patches

Microsoft uses the Microsoft Active Protections Program, known as MAPP, to manage security fixes. The programme is managed by the Microsoft Security Response Center and shares early details with partners before Patch Tuesday, the company's monthly patch release day. MAPP is now 18 years old. A related process, the Software and Services Incident Response Plan, dates to 2004.
Who is affected and what is at risk
Most new reports come from researchers outside the usual security community, and several people often report the same flaw at once. The bugs are generally less severe, with the average CVSS score falling from about 7.8 to 7.4 as of September. Yet each report still needs human review. Microsoft has also issued Windows patches for some Office flaws before Mac versions, leaving Apple users exposed while they wait.
What Microsoft says
Chief executive Satya Nadella said at the New York Times DealBook Summit that a recent Microsoft 365 outage resulted from a latent bug rather than a direct AI attack. He also told CNBC that applying many patches to older infrastructure can make systems less stable. He warned that finding and fixing bugs with AI creates more patching, which brings instability.
Sources
Spotted a mistake? See our corrections policy. Read how News by OnDevice works in our editorial policy.



