Executive Summary
Hidden instructions in a Word document can make Microsoft 365 Copilot rewrite figures in a report and copy the same instructions into the new file, which then reinfects later drafting sessions; disclosed 144 days after reporting.
What Happened
Hidden instructions in a Word document can make Microsoft 365 Copilot rewrite figures in a report, then copy the same instructions into the finished file. Håkon Måløy disclosed the technique on July 28, 144 days after reporting it to Microsoft. In his proof of concept, the internally generated file triggered the same behavior when it was used in a second Copilot drafting session. Måløy's
Affected Products
- Microsoft 365 Copilot for Word
Vulnerability / Attack Technique
Self-propagating prompt injection in documents
Technical Mechanism
Copilot follows hidden instructions in a source document, alters figures, and writes the instructions into its output, spreading them.
AI Security Category
LLM01:2025 Prompt Injection
Part of the Prompt Injection topic.
Relevant Frameworks
RealGround Analysis
Silent changes to report figures are an integrity risk, not only a leak. RealGround analysis: verify numbers in AI-drafted documents against sources.
Recommended Controls
- Check figures in Copilot-drafted documents against source data
- Scan documents for hidden text before use with Copilot
Related RealGround Service
AI Agent Business Logic Audit · or start with the free AI Agent Security Scanner.
Sources
thehackernews.com: https://thehackernews.com/2026/07/microsoft-copilot-for-word-can-copy.html
Summary and analysis written by RealGround from the source above; last updated 2026-10-18.