Attack ChainAIDerived
Hypothetical attack flow derived from CVE metadata – not a confirmed exploit chain
-
AccessCraft malicious prompt with embedded commands
-
DeliverySend to M365 Copilot via network
-
ExploitCopilot parses command injection
-
ExecutionExecute unauthorized information disclosure
-
ImpactExfiltrate sensitive data
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 | M365 Copilot must be accessible to attacker; user must interact with crafted prompt (UI:R); no authentication required; default or standard Copilot configurations vulnerable to command injection payloads. |
| Impact | Despite the high CVSS score of 7.1, the real-world risk appears moderate. The CVSS vector (AV:N/AC:L/PR:N/UI:R) indicates remote exploitation is possible without authentication but requires user interaction. Key risk factors: No EPSS score suggests low exploitation probability, not listed in CISA KEV indicating no known active exploitation, no public POC available, and patches are already released. The 'High' confidentiality impact could expose sensitive business data through Copilot, but the user interaction requirement and lack of exploitation evidence reduce immediate risk. This is likely a high-CVSS-but-moderate-real-risk issue that should be patched during normal maintenance windows rather than emergency response. |
| Exploit ScenarioDerived | An attacker could craft malicious content (document, email, or webpage) containing specially formatted AI command injection payloads. When a user interacts with this content and uses Copilot features, the malicious prompts could manipulate the AI to disclose information it shouldn't access, potentially revealing sensitive business data, personal information, or internal documents indexed by Copilot. The attack requires the victim to actively engage with Copilot while viewing the malicious content, limiting automation potential. No public POC exists, suggesting the technical details remain closely held. |
Get an e-mail when this CVE is exploited, patched or re-scored. Free account.
Alert meAnalysisAIv1 – Mar 13, 2026
CVE-2026-26133 is an AI command injection vulnerability in Microsoft 365 Copilot and multiple Microsoft mobile/desktop applications that allows remote attackers to disclose sensitive information through crafted AI prompts. The vulnerability affects numerous Microsoft products across iOS, Android, and macOS platforms, requires user interaction, and has a patch available from Microsoft with no current evidence of active exploitation (not in KEV).
RemediationAI
Microsoft has released patches for all affected products - update to the latest versions specified in the EUVD data. Primary remediation: Apply vendor patches through standard app store update mechanisms for mobile apps or Microsoft Update for desktop applications. Vendor advisory available at: https://msrc.microsoft.com/update-guide/vulnerability/CVE-2026-26133. As an interim mitigation, organizations could restrict Copilot functionality or implement additional monitoring for suspicious AI prompt patterns until patching is complete.
Affected VersionsENISA
Affected Packages
Source: OSV (Open Source Vulnerabilities)
Affects your stack
Affects 40 packages in microsoft, euvd
Checking this CVE against your own stacks and servers is part of the Basic plan and up.
Technical ContextAI · show details
This vulnerability represents a new class of AI-specific security issues - command injection through AI prompts in Microsoft's Copilot integration. The affected products span Microsoft's entire mobile ecosystem including M365 Copilot, Edge, Excel, Word, PowerPoint, OneNote, Outlook, Teams, Loop, and PowerBI across Android, iOS, and macOS platforms. While no CWE is assigned, AI command injection typically involves manipulating AI system prompts to bypass intended restrictions and access unauthorized data. The vulnerability specifically targets the Copilot AI assistant functionality integrated across these applications.
Know when Microsoft 365 Copilot For Android 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:R/S:U/C:H/I:L/A:N
Primary rating from NVD · only source for this CVE.
CVSS VectorNVD
Lifecycle Timeline
2DescriptionCVE.org
AI command injection in M365 Copilot allows an unauthorized attacker to disclose information over a network.
Executive ViewAI
CVE-2026-26133 is a high-severity AI command injection vulnerability affecting Microsoft 365 Copilot and multiple Microsoft applications on iOS, Android, and macOS that could allow attackers to extract sensitive information through crafted prompts. While the exploit requires user interaction and has no current evidence of active exploitation, the widespread impact across Microsoft's ecosystem and the sensitive nature of data accessible through these applications necessitates prompt patching to minimize organizational risk.
Apply vendor patch.
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.
Remote code execution in the Microsoft 365 Copilot mobile apps for Android and iOS lets an unauthenticated attacker run
Spoofing via improper access control in Microsoft 365 Copilot for Android (versions prior to 16.0.19822.20190) enables a
EUVD-2026-12111
NVD
CVE.org
EPSS
VulDB
GitHub PoC
Exploit-DB