Kevin Mandia exposes the AI cyber tsunami at a16z
By Bram de Vries · Reporting from Amsterdam ·
On the a16z podcast, Kevin Mandia argues that autonomous AI cyberattacks require enterprises to abandon manual defense for machine-speed prevention.
When Kevin Mandia looks at the coming wave of machine-speed attacks, he does not see a distant regulatory abstraction; he sees a tsunami that dwarfs human capacity. On the a16z podcast episode Building Cyber Defense for the Agentic Era, Mandia and general partner David George discussed how Armadin has found over 90 zero-day vulnerabilities in production at Fortune 500 companies since January alone. If open models can probe thousands of paths simultaneously in microseconds, the traditional architecture of enterprise defense is finished. ## The Death of Manual Security and the Rise of Autonomous OffenseFor thirty years, security meant playing catch-up behind human adversaries who moved along single, deliberate paths. As Mandia noted, AI changes the economics of cyberattacks by replacing the sniper round with a drone swarm that exhausts tokens and tests every route continuously. Traditional penetration testing is merely a hygiene check that misses logic flaws in custom applications. When Armadin deploys its offensive AI to map every service and route, it operates at a scale that leaves human defenders blinking in the dark. You cannot defend a modern enterprise with spreadsheets and manual incident response when the adversary is operating at machine speed. ## The False Comfort of Open-Closed DistinctionsThere is a comforting delusion in Brussels and among cautious regulators that we can simply slow down the models or legislate safety through bureaucratic delay. Mandia dismantles this by pointing out that open-weight models and closed frontier systems perform essentially the same when pushed through full attack kill chains. The differentiation between open and closed shrinks rapidly once models are post-trained for specific tactical execution. State-aided caution protects yesterday's compliance consultants, but it leaves commercial enterprises exposed to anonymous attackers who face zero friction in acquiring GPU capacity. Markets reward speed and efficacy, not committees drafting directives about cybersecurity risk. ## Autonomous Triage as the New True NorthThe ultimate destination for enterprise security is not more alerts; it is effective autonomous response. Mandia explains that Armadin Blue must create compensating controls at speed because humans cannot intervene in the detect-and-respond loop without being hopelessly outpaced. By working alongside established defense platforms like CrowdStrike and Palo Alto Networks, autonomous systems can apply a field-dressing tourniquet before an intrusion takes root. Enterprises that devote engineers to fortifying their own walls are making the hard trade required to survive a digital market where the first company to industrialize prevention wins. Prosperity in the digital age belongs to those who build the automated force field, not those who wait for permission to defend themselves.