Microsoft Copilot Chat
CVE-2025-59286
CRITICAL
Attack ChainAIDerived
Hypothetical attack flow derived from CVE metadata – not a confirmed exploit chain
-
AccessReach Copilot Chat over network
-
DeliverySubmit crafted input with command elements
-
ExploitService fails to neutralize special elements
-
ExecutionInjected command processed across trust boundary
-
ImpactSensitive information disclosed to attacker
Vulnerability Assessment
| Weakness |
CWE
OWASP Top 10:2025
A05:2025 - Injection
ASVS 5.0
V1 - Encoding and Sanitization Command injection prevention
MITRE ATT&CK
1 technique across 1 tactic: Execution The technique list with ATT&CK links and the kill-chain view is part of the Basic plan and up. |
| Exploitation Conditions | Exploitation targets Microsoft 365 Copilot Chat specifically; the attacker must be able to send input to the Copilot Chat service over the network and supply content containing unneutralized special command elements that the service interprets. Per the CVSS vector (AV:N/AC:L/PR:N/UI:N), no authentication, no privileges, and no user interaction are required, and complexity is low, so there is no special configuration prerequisite stated. The main limiting factor is that this is a Microsoft-hosted cloud service: Microsoft controls the single deployment, so once the vendor remediates server-side the window closes for all tenants, and individual customers cannot be targeted via a locally misconfigured instance. The scope change (S:C) indicates the disclosed information can extend beyond the component the attacker directly interacts with. |
| Impact | Signals are mixed and point to a serious but not yet mass-exploited issue. The CVSS 3.1 base score of 9.3 (AV:N/AC:L/PR:N/UI:N/S:C/C:H/I:L/A:N) describes a network-reachable, low-complexity, unauthenticated flaw with a scope change and high confidentiality impact - a textbook high-priority profile. However, EPSS is only 0.57% (45th percentile), the CVE is NOT on the CISA KEV list, and no public PoC is referenced, so there is no evidence of active or imminent widespread exploitation. There is also a notable internal tension worth flagging: the CWE is CWE-77 ('command injection') and the title says command injection, yet the impact metrics are C:H/I:L/A:N (information disclosure with only limited integrity impact and no availability impact), which is atypical for classic OS command injection and is more consistent with an AI/prompt-injection-style flaw in the Copilot service that leaks data across a trust boundary (hence S:C) - verify the true technical nature with Microsoft before treating it as remote code execution. Critically, 365 Copilot Chat is a Microsoft-operated cloud/SaaS service, so the fix is almost certainly already deployed server-side by Microsoft and customers likely have no patching action; this materially lowers the practical, ongoing risk to a given tenant despite the high nominal score. Net assessment: a genuinely high-severity design/injection flaw that warranted an urgent vendor fix, but current real-world risk to customers is moderated by SaaS remediation and the absence of KEV/PoC/high-EPSS exploitation signals. |
| Exploit ScenarioDerived | An unauthenticated remote attacker crafts input containing special command elements that the 365 Copilot Chat service fails to neutralize, causing the service to process the injected instructions and return information the attacker should not be able to access. Because the scope is changed (S:C), the disclosed data can cross a trust boundary (e.g., other context or tenant resources) rather than being limited to the attacker's own session. No public exploit code is identified at time of analysis, and the attack requires no privileges or user interaction per the CVSS vector. |
Get an e-mail when this CVE is exploited, patched or re-scored. Free account.
Alert meAnalysisAIv1 – Oct 08, 2026
Crafted command input submitted to Microsoft 365 Copilot Chat can cause the hardened service to disclose sensitive information across a trust boundary, because the service fails to neutralize special command elements before interpreting user-supplied content (CWE-77). The flaw is reachable over the network with no authentication, no privileges, and no user interaction (CVSS 9.3, CVSS:3.1/AV:N/AC:L/PR:N/UI:N/S:C/C:H/I:L/A:N), so any organization whose users rely on the hosted Copilot Chat service is exposed. No public exploit code and no confirmed active exploitation have been identified at time of analysis, and EPSS is low at 0.57% (45th percentile); because this is a Microsoft-operated cloud service, the exposure window closes for all tenants once Microsoft completes server-side remediation.
RemediationAI
Because Copilot Chat is a Microsoft-operated cloud service, remediation is applied server-side by Microsoft rather than by customers installing a patch; consult and track the vendor advisory at https://msrc.microsoft.com/update-guide/vulnerability/CVE-2025-59286 (MSRC update guide) for the remediation status and any tenant-side action items. Until server-side remediation is confirmed, compensating controls include restricting or disabling access to Copilot Chat for non-essential users via Microsoft 365 admin controls, which reduces the attack surface but removes assistant functionality for those users; limiting the data connectors, Graph permissions, and indexed content that Copilot Chat can reach, which lowers the value of any disclosed information but may degrade answer quality; applying sensitivity labels and DLP policies so that highly sensitive content is not retrievable by the assistant, which introduces policy-management overhead; and enabling Microsoft 365 audit logging and prompt/response monitoring so that anomalous command-like inputs can be detected and investigated, which does not prevent exploitation but shortens detection time. Note that no public exploit code was identified at time of analysis, so prioritization should be driven by the high CVSS score (9.3) and the breadth of the hosted service rather than by observed exploitation.
Affects your stack
Find out whether the stacks and servers you run are on an affected version.
Checking this CVE against your own stacks and servers is part of the Basic plan and up.
Technical ContextAI · show details
The root cause is CWE-77 command injection: special elements embedded in attacker-supplied input are not properly neutralized before being interpreted by the Copilot Chat service, allowing an attacker's content to be treated as commands rather than inert data. The affected component is identified by CPE cpe:2.3:a:microsoft:365_copilot_chat:-:*:*:*:*:*:*:*, i.e., the hosted Microsoft 365 Copilot Chat service rather than an installable on-premises binary. The CVSS scope-change flag (S:C) indicates that information disclosed through the flaw can affect resources beyond the component the attacker directly interacts with, which for an LLM-backed assistant implies potential exposure of data reachable by the service's connectors and context. Integrity impact is low (I:L) and availability impact is none (A:N), consistent with an information-disclosure-centric primitive rather than destructive code execution. The practical limiting factor is deployment topology: Microsoft controls the single production instance, so there is no locally misconfigured tenant instance for an attacker to target and no customer-side patch to deploy.
Know when 365 Copilot Chat gets its next vulnerability
- Watch products and get an email when something new is published
- Check this CVE against your own stack and servers
- Exploit and PoC references, threat intelligence from 22 sources
No credit card · 7-day full trial
Severity by source
AV:N/AC:L/PR:N/UI:N/S:C/C:H/I:L/A:N
Network-reachable, unauthenticated, low-complexity injection with a cross-boundary (S:C) high confidentiality leak and only limited integrity, no availability impact.
Primary rating from Vendor (microsoft).
CVSS VectorVendor: microsoft
DescriptionCVE.org
Improper neutralization of special elements used in a command ('command injection') in Copilot allows an unauthorized attacker to disclose information over a network.
Executive ViewAI
Microsoft 365 Copilot Chat contains a critical unauthenticated information-disclosure flaw that can allow crafted command inputs to cross trust boundaries and expose sensitive data. Because the service is Microsoft-operated, any organization whose users rely on Copilot Chat is potentially affected, but remediation is expected server-side across all tenants. Leadership should treat this as a high-priority data-protection risk and require immediate visibility, usage controls, and vendor confirmation of remediation.
No vendor-released patch identified at time of analysis. Interim mitigations include restricting or disabling Microsoft 365 Copilot Chat for sensitive user groups, enforcing data loss prevention and sensitivity labels, monitoring Copilot audit logs for anomalous cross-boundary data access, applying least-privilege and conditional access policies, and engaging Microsoft support for server-side remediation timelines.
Threat Intelligence Overview
How this decision was reached
The engine reads each signal below and applies a fixed list of rules from top to bottom. The first rule that matches sets the recommended action – like a statement, every line is shown.
What every action level means
SSVC AssessmentENISA
Priority Score
Attack Technique
OS command injection occurs when an application passes unsanitized user input directly into system shell commands.
Affected Products (CPE)CVE.org
More in 365 Copilot Chat
View allImproper neutralization of special elements (CWE-77 command injection) in Microsoft 365 Copilot Chat allows an unauthent
Unauthenticated command injection (CWE-77) in Microsoft 365 Copilot's Business Chat allows an attacker who can supply co
Microsoft 365 Copilot BizChat Information Disclosure Vulnerability. Rated high severity (CVSS 8.2), this vulnerability i
Microsoft 365 Copilot BizChat Information Disclosure Vulnerability. Rated medium severity (CVSS 6.5), this vulnerability
NVD
CVE.org
EPSS
GitHub PoC
Exploit-DB