What the August figures measure
Fewer than half of the Windows devices in an enterprise sample had installed the August security update two weeks after Microsoft released it. Four weeks out, about a quarter still lacked it. Nuno Costa, who oversees Windows servicing and delivery, reported the figures on November 5. Microsoft did not disclose the sample size or how devices were selected. The figures describe installation in that group; they cannot establish a rate for all Windows customers, or tell us whether any of the devices came under attack.
Two numbers from different clocks
In the same announcement, Microsoft said its October Windows release addressed more than 1,100 vulnerabilities. It described using AI to help engineers investigate findings, prepare fixes across versions and choose tests. Engineers remain responsible for release decisions. The October count and the August deployment figures concern different releases and measure different things: one counts problems addressed in code; the other follows devices after a fix is available. Neither shows that AI has shortened the time a customer waits for protection.
What a service provider can actually attest to
For a managed service provider reporting to a customer, the distinction changes the meaning of a monthly patch report. If two customers approve the same cumulative Windows update on the same day, their approval records look identical even if one fleet spends another month on the old build. The underlying vulnerability list cannot distinguish the two customers either. The evidence that separates them is when their devices installed the update and which devices have yet to report. An endpoint that has gone silent is an unknown, not proof of a failed installation. Microsoft’s sample makes the business consequence visible: by the time the next monthly release approaches, the previous rollout can still be unfinished.
Sources & further reading
Our practical suggestions are editorial guidance. Verify applicability against current source and vendor instructions.
