The $6B Binance heist, Meta’s AI bug tests and looming quantum deadlines

IBM's Security Intelligence podcast covers the $6.34B Binance heist, Meta's AI-driven "just-in-time" bug testing, and quantum-safe migration planning 1:06.

Key Takeaways:
• Binance lost ~$6.34B in a months-long heist starting with an infected macOS package manager and a fake job-interview assignment; attackers used malware and internal AI tools to expand access 1:41-2:21
• The attack succeeded because attackers researched both technical systems and business processes—attackers don't operate in silos, but defenders do 2:31-4:45
• Mitigations: add "friction" at key hops, assume devices are burned (revoke tokens, keys, sessions), treat internal AI tools as privileged users, and educate high-privilege staff 7:12-9:38
• Meta's JIT testing uses LLMs to generate bug tests at pull-request time, distinguishing context-unaware from context-aware workflows 13:57-14:57
• Panelists favor "human on the loop" guardrails, cautioning against early autopilot since AI amplifies whatever exists in an organization 16:16-21:34
• With the 2030 post-quantum deadline, "harvest now, decrypt later" threatens long-lived data; treat cryptography as a lifecycle and start early—it's an execution challenge more than a math one 23:02-31:35

The recurring theme: automate containment, embrace zero trust, and start long migrations early.

Sources:

• 1:41-2:21 Binance heist timeline and attack chain
• 2:31-4:45 Attacker sophistication vs. defender silos
• 7:12-9:38 Zero trust friction and containment advice
• 13:57-14:57 Meta's JIT LLM-based testing
• 16:16-21:34 Human-on-the-loop debate
• 23:02-31:35 Post-quantum migration urgency