Cloud Security Podcast by Google EP243 Email Security in the AI Age: An Epic 2025 Arms Race Begins
Sep 15, 2025
Cy Khormaee, CEO of Aegis AI, and Ryan Luo, CTO of Aegis AI, discuss the future of email security amid escalating AI threats. They reveal that by 2025, traditional defenses may falter against sophisticated cyber attacks. The conversation emphasizes the importance of adaptive models and the crucial role of good AI in combating bad AI. They share insights from their experiences at Google and stress the need for innovative strategies to counter personalized phishing tactics. Best practices for using large language models are also highlighted.
AI Snips
Chapters
Transcript
Episode notes
AI Supercharges Email Threats
- Email threats exploded and became far more sophisticated with AI enabling higher volume and novel attack types.
- AI powers both advanced persistent-type attacks and brand-new evasion techniques that defeat old defenses.
Cat-and-Mouse Is Too Slow
- Attacks now iterate in minutes or seconds creating bespoke, one-off campaigns.
- Defenders must detect zero-days preemptively because reactive rule-writing is too slow.
Use LLMs Selectively For Intent
- Use large language models selectively for deep language and intent understanding rather than blanket rule systems.
- Include only essential context in queries to control cost and reduce unreliability.