Security Weekly Podcast Network (Video) Typing, textGrain, NetScaler, Fortinet, Copilot, LibreOffice, Dots, Aaran Leyland - SWN #622
Oct 6, 2026
Aaron Leyland, a cybersecurity commentator focused on phishing, identity security, and authentication, joins a wide-ranging discussion. Topics include AI-assisted bank attacks, TextGrain watermarking, recurring NetScaler and FortiMail flaws, Copilot prompt injection, dangerous LibreOffice files, and privacy risks from personal AI assistants. Leyland also explores relationship-based phishing, session theft, and why passkeys matter.
AI Snips
Chapters
Books
Transcript
Episode notes
Attackers Exploit The Systems Around Critical Infrastructure
- Attackers targeted bank support systems rather than heavily defended core banking systems.
- Distributed data lets attackers assemble sensitive profiles across multiple overlooked systems, much like an SQL join.
Map Data And AI Activity Before Attackers Do
- Map data flows, test controls, and log what AI agents access and do.
- Connect security, privacy, and AI governance before an incident exposes gaps.
TextGrain Watermarks May Fail Under Editing
- OpenAI’s TextGrain watermark changes word choices to create a detectable statistical pattern.
- Editing can erase it, and mathematical or scientific writing offers too little wording flexibility.









