The Library of Minds Cybersecurity Investor: Your OpenClaw Will Betray You with Enrique Salem (Partner, Bain Capital Ventures)
24 snips
Apr 23, 2026 Enrique Salem, former CEO of Symantec and current Bain Capital Ventures partner with decades in cybersecurity and investing. He explains how AI agents amplify personalized phishing and inherit risky credentials. He warns about model poisoning, agents leaking sensitive data, and why security must prioritize what matters rather than trying to protect everything.
AI Snips
Chapters
Transcript
Episode notes
Agents Replace Judgment With Risky Automation
- Agents inherit credentials and can act without human judgment, creating subtle risks.
- He gave the example of an agent reposting earnings press releases before earnings day because it lacks the human discretion to wait.
Assume Agents Will Be Poisoned
- Expect attackers to poison agents and design controls assuming agents will be manipulated.
- Salem predicts attacks like false calendar entries that trick executives into approving payments or meetings.
Model Poisoning Outweighs Vulnerability Risk
- Model poisoning is a bigger long-term worry than traditional software vulnerabilities.
- Salem reasons defenders can limit blast radius of code exploits, but poisoned model weights can subtly undermine behavior at scale.