Artificial intelligence is rapidly shifting the balance of power in cybersecurity by allowing attackers to discover software vulnerabilities and deploy exploits in real time. According to Alex Stamos, chief security officer at Cognition and a lecturer at Stanford University's Cyber Policy Center, modern open-weight AI models can now automate substantial portions of the "kill chain"—the sequence of steps hackers execute to penetrate a target—challenging long-standing defensive strategies.
Traditionally, cyber operations required human specialists to spend weeks or months examining code, finding flaws, and developing tailored exploit tools before manually launching an attack against selected targets. With AI, that dynamic has inverted. Automated systems can now scan target networks, identify unpatched bugs on the fly, and write functional exploit code simultaneously, drastically reducing the effort and time required to breach an organization.
The shift arrives amid a wave of cascading corporate intrusions where compromised services are linked together to amplify damage. Recent supply-chain incidents illustrate how attackers can compromise one infrastructure provider, steal credentials to access developer platforms, and subsequently expose thousands of downstream companies. Open-source offensive tools powered by large language models have already been linked to breaches targeting banks and hundreds of commercial enterprises.
Stamos compared the current disruption to the chaotic expansion of the online frontier in the 1990s, warning that the industry faces a transitional "dark age" as security teams scramble to adapt. While noting he is not pessimistic in the long run, he emphasized that defenders face mounting pressure to accelerate vulnerability detection before automated offensive tools outpace conventional safeguards entirely.