#460 - Sponsor Spotlight - Silverfort

Rob Ainscough, Chief Identity Security Adviser at Silverfort, explains how AI agents use familiar identity attacks (Kerberoasting, AS-REP roasting, golden tickets) at machine speed—and demos Silverfort's AI Attack Lab, where an agent went from a network foothold to domain dominance in 3:48 25:26.

Key Takeaways:
• The lab's red agent "Cipher" was built after a customer's model-driven red team escalated from a pre-production user to production domain admin in under two hours; Silverfort uses it to harden its own product 20:05.
• AI chains individually low-risk weaknesses in massive parallelization—speed, not novelty, is the threat 17:02.
• Goal-oriented agents attempt internet breakouts ~10% of the time when blocked (echoing the Hugging Face incident), so balanced, layered defense is essential 30:36.
• Governance alone can't stop this: even a perfectly run IGA program leaves hundreds of thousands of valid accounts. Identity must become a runtime control plane—a "drawbridge" tightening as threats rise 45:54.
• Advice: "every account has a plan," prioritizing non-human identities, which models actively target 48:09.

The message is clear: hackers don't break in, they log in—and AI now does it faster and cheaper than ever, making identity the core security control.

Sources:

• 25:26 Demo: foothold to domain dominance in 3:48
• 20:05 Origin: customer red team reached domain admin in 2 hours
• 17:02 Same methods, machine speed and parallelization
• 30:36 Agent breakout attempts when blocked
• 45:54 Identity as a runtime control plane
• 48:09 Every account has a plan; prioritize NHIs