Ami Luttwak shows why AI attacks demand action over bloat
By Grant Colby · Reporting from Amarillo ·
On the "a16z" podcast, Ami Luttwak warned that AI models collapse attack times to minutes, proving corporate compliance bloat cannot replace active deterrence.
The time required for an adversary to breach a private company's network and drain its critical assets has collapsed from months to minutes. On the "a16z" podcast, Wiz co-founder and CTO Ami Luttwak and General Partner Zane Lackey laid out what security professionals call "Black October." Open-source artificial intelligence models have caught up to offensive capabilities. That gives routine intruders the speed and expertise of nation-state threat actors.
The Digital Flight Line Is Under Fire
On the podcast, Luttwak explained that open-source AI models now possess the capability to independently scan external code interfaces. They discover zero-day vulnerabilities and execute full breaches before a response team can evaluate an alert. Zane Lackey warned that this operational tempo threatens to push the private sector back into the wild west of earlier decades. Unauthenticated worms infected systems faster than human operators could apply software updates. Luttwak pointed to a critical flaw in PHP, the server-side code powering much of the web. While part of Wiz's client base patched quickly, legacy systems and Linux distributions remain exposed because official patches take weeks or months to trickle down. Meanwhile, AI development platforms are targeted specifically so attackers can steal access keys and run up illicit charges.
Compliance Bureaucracy Won't Stop an Aggressor
Faced with this mounting threat, corporate boardrooms will be tempted to handle the crisis the way Washington handles every national challenge. They will inflate security budgets, hire more compliance officers, and churn out regulatory reports. Defenders argue that rigid mandates force sprawling enterprises to maintain basic asset inventories. They say compliance establishes legal accountability across legacy infrastructure. But as Luttwak admitted during the conversation, most of what corporate security teams do today is not actual security. It is administrative process and static checking that creates an illusion of safety while leaving the perimeter wide open. I balanced payrolls under Carter-era inflation and flew C-130s through Desert Storm. Clipboards and compliance binders never secured a flight line. Defensive deterrence requires hard operational capability, not bureaucratic paper trails.
Deterrence Demands Continuous Offense
Peace through strength is not just a strategic doctrine for military commanders. It is the sole surviving playbook for American enterprise. Luttwak argued that the only way to defend an organization today is through automated "red agents." These offensive AI systems continuously attack a firm's public-facing interfaces from the outside to locate vulnerabilities before an enemy does. Luttwak insisted that any digital infrastructure incapable of withstanding this relentless automated stress should simply be disconnected from the public internet. American enterprise cannot rely on Washington regulators or corporate hand-wringing. Open-source tools hand supercomputer-level weapon systems to hostiles abroad and criminals at home. Free enterprise will remain secure on Main Street only when business leaders abandon useless administrative overhead. They must confront realities on the ground and build digital defenses capable of out-fighting the enemy in real time.