Microsoft Flags AI-Driven Phishing: LLM-Crafted SVG Files Outsmart Email Security
Summary: Microsoft reported a phishing campaign, detected on August 28, 2025 and aimed primarily at U.S.-based organizations, that hid its payload in an SVG file disguised as a PDF file-sharing notification and appeared to be built with a large language model. The messages came from a compromised business email account, used matching sender and recipient addresses with targets hidden in BCC, and the SVG redirected victims through a CAPTCHA to a fake login page to harvest credentials.
What's compelling: Microsoft's Security Copilot judged the code too complex, verbose, and impractical for a human to write from scratch, and the SVG format's invisible elements, encoded attributes, and delayed script execution are built to defeat static analysis and sandboxing. Microsoft said the campaign was limited in scope and effectively blocked, but that similar techniques are increasingly being used by a range of threat actors.
Read the original at The Hacker News ↗
- Topics
- Detection & Response
- Related
- YARA-X 1.23.0 Release, (Sun, Nov 15th) Cyber Risks · Nov 15, 2026Inside Falcon AI Gateway: One Control Point for AI Traffic Cyber Risks · Nov 13, 2026Macfinger Click Fix campaign sends Down TheDRAIN infostealer for macOS, (Fri, Nov 13th) Cyber Risks · Nov 13, 2026