AI tools accelerate cyberattacks, security chief warns of expanding breach chain
Cognition's chief security officer says open‑source AI models are automating the hack cycle, leading to breaches at Binance and a cascade affecting thousands of firms.

On Nov. 7, Alex Stamos, chief security officer at Cognition and veteran cybersecurity analyst, told The Verge that artificial‑intelligence‑driven hacking tools are powering a surge of attacks that have already compromised dozens of organizations, including the cryptocurrency exchange Binance and a chain of incidents that reached roughly five thousand downstream companies.
Stamos outlined a series of recent events that illustrate the scale of the problem. Vulnerabilities in widely used software such as PHP and Node were exploited, and a succession of breaches at AI‑focused “neocloud” providers led to the theft of API keys from the service Fireworks. Those keys were used to infiltrate the development platform Cursor, where a malicious plugin was deployed to about five thousand client organizations, creating potential breaches on a massive scale.
The core of the escalation, according to Stamos, is the ability of open‑weight AI models to automate the traditional cyber‑kill chain. These models can scan target systems, discover vulnerabilities, generate exploits and launch attacks in real time, eliminating the lengthy manual research that previously preceded large‑scale breaches. The open‑source tool known as ARTEX, hosted on GitHub and powered by a model comparable to GLM‑5. 3, exemplifies this capability.
Both private cybercriminal groups and state‑backed actors such as Russia and North Korea are reported to be leveraging these AI tools. The accessibility of open‑source code means that techniques that once required specialized expertise can now be downloaded and deployed by anyone with modest technical skill, widening the pool of potential attackers.
Stamos said the multibillion‑dollar cybersecurity industry is struggling to keep pace. Traditional monthly security awareness campaigns – urging password changes and the use of password managers – have proved insufficient as the sector operates at a heightened alert level he likened to DEFCON 2. He noted that roughly two hundred additional companies were hit this week alone by ARTEX‑enabled attacks, underscoring the breadth of the emerging threat.
He compared the current environment to the early internet era of the 1990s, when Windows vulnerabilities left systems exposed, but stressed that today nearly every critical service runs online, and AI now automates exploitation that once required weeks of manual effort.
Looking ahead, Stamos warned of a looming “dark age” as firms scramble to develop defenses against AI‑powered threats. While Cognition does not provide incident‑response services, it assists affected firms in patching vulnerabilities after breaches and emphasizes the need for new security paradigms to contain the rising tide of AI‑driven attacks.
Sources
- AI broke the cybersecurity world. What’s next?AI | The Verge
