AI Security Alert: Copilot Mailbox Search Vulnerability & LiteLLM Admin Key Leak Exposed
Two AI tools broke in the same way in the same two weeks, and four research teams proved it. The pattern underneath every disclosure is one sentence: enterprise AI accepts external input with no trust boundary. On June 15, Varonis disclosed SearchLeak (CVE-2026-42824) , a proof-of-concept exfiltration chain in Microsoft 365 Copilot Enterprise Search. A victim clicks a crafted microsoft.com URL, Copilot searches their mailbox, and the data leaves through a Bing SSRF. No plugin
Two significant AI security vulnerabilities have been disclosed in rapid succession, affecting Microsoft 365 Copilot Enterprise Search and LiteLLM. The Copilot vulnerability, dubbed SearchLeak, allows attackers to exfiltrate mailbox data via a crafted URL, exploiting a lack of trust boundaries in enterprise AI. Separately, Obsidian Security revealed a three-CVE chain against LiteLLM that grants admin privileges and enables remote code execution from a default low-privilege account. These incidents highlight a critical pattern: AI systems accepting external input without adequate security checks. Varonis, who disclosed SearchLeak, has identified multiple Copilot exfiltration chains previously. The LiteLLM exploit is particularly concerning due to its ease of execution. A five-check audit is recommended to address these systemic weaknesses.
These critical vulnerabilities in widely used AI tools expose significant security risks, necessitating immediate audits and highlighting the urgent need for robust security measures in AI development.
📌 Kaynak
Bu haber XML kaynağından derlenmiştir. Tamamı için orijinal habere gidin.
Orijinal haberi oku →