AI agents breached about 100 companies after hitting over 1,200 websites

The incident, first reported by Reuters and discovered by cybersecurity researchers, illustrates how AI agents can operate autonomously at scale. While the human hacker did not execute the attack directly, they had to pay for the AI tools, which cost about $5,000 in tokens from DeepSeek and Anthropic's Claude Code models. This works out to roughly $50 per successful breach, a figure that experts say fundamentally changes the economics of exploitation.

Jessica Ji, a senior research analyst at Georgetown's Center for Security and Emerging Technology (CSET), told Fortune that autonomous attacks are becoming more common because AI models have become meaningfully better over the past year. Specifically, improved coding skills have led to a dramatic increase in their ability to perform cybersecurity-relevant tasks like analyzing large volumes of code. Additionally, models have become more agentic, meaning they can use other software to complete tasks rather than merely outputting text. "These AI systems are now both more capable of hacking and are able to act more autonomously," Ji said.

Varun Chandrasekaran, an assistant professor at the University of Illinois, noted that LLM-powered agents can run all day and night without stopping, at a fraction of the cost of human attackers. The debate over open-weight models also plays a role: while Anthropic CEO Dario Amodei has warned that open models pose a risk of widespread cyber attacks and suggested government restrictions, others argue open models benefit defenders. For example, when Hugging Face was attacked by an agent in July, it switched to an open model, GLM 5.2 from China-based Z.ai, after a closed model refused its request due to security guardrails.