CyberNotableConfirmedDeveloping
9.2
Security researcher identifies prompt injection vulnerability in Microsoft Copilot
vx-undergroundLO·US·about 22 hours ago
Researcher Håkon Måløy demonstrated vulnerabilities in Microsoft Copilot that could theoretically enable the creation of self-replicating AI worms within Office documents. The exploit relies on prompt injection techniques to execute unauthorized instructions, though the practical feasibility of large-scale deployment remains unverified.