CVE-2026-78501 is a high severity (CVSS 8.6) information disclosure flaw in Microsoft 365 Copilot Business Chat. Microsoft lists no customer action for this service. Microsoft has not marked it as exploited at release, and the CISA KEV flag is false. Check the MSRC entry for your exact build.
What's affected
Microsoft 365 Copilot Business Chat, a cloud service. Microsoft's data for this CVE lists the customer action required as No, and shows no downloadable update, which fits a service that Microsoft updates itself.
The record does not list versions. Microsoft's September 2026 security update data, which we pulled from the Microsoft Security Response Center API, marks the issue as not publicly disclosed and not exploited at release. The record's KEV flag is false, so there is no sign of exploitation in the wild.
How to check if you're exposed
There is nothing to install for a service Microsoft operates, so there is no version to check. What you can check is your own exposure: who in your organization uses the service, which data it can reach, and what your tenant's access and audit settings look like.
Read the entry in the Microsoft update guide for the current status, since Microsoft can change the notes after release. If your tenant shows any related advisory or message center post, read it.
What to do now
- Open the Microsoft update guide entry for this CVE and confirm the status and any customer guidance.
- Confirm in your admin portal that the service is on its normal update channel, since the fix is applied by Microsoft.
- Review who can use the affected service and what data it can reach, and tighten access where it is broader than needed.
- Review audit logs for unusual activity in the period around the disclosure, and keep logging enabled.
Why it matters
The weakness is CWE-77, command injection. The record lists the class as improper neutralization of special elements used in a command, along with a second weakness, CWE-923, improper restriction of communication channel to intended endpoints. In assistants that read data on a user's behalf, hostile text can be treated as an instruction. Microsoft's description gives no mechanism, so we say no more than that the result is disclosure of information across a scope boundary.
The published description is a single sentence, and Microsoft does not add technical detail for most of these issues. That is typical, and it is the reason to judge risk by exposure: who can reach the vulnerable component, and how fast you can patch it. The vector has no privileges and no user interaction, with a changed scope, and the impact is on confidentiality only.
Quick check
Do I need to patch anything?
Microsoft's data lists customer action as not required for this service.
Does this mean data was leaked?
Neither the record nor Microsoft's data says it was exploited.
Sources
Last reviewed: 18 October 2026.
