Late last year, Anthropic disclosed something unprecedented: state-sponsored Chinese hackers had used its AI technology to infiltrate the computer systems of roughly 30 companies and government agencies worldwide. Human hackers handled about 10 to 20 percent of the work. The AI agent did the rest.

That was the first known case of a cyberattack driven largely by an AI agent. Five months later, it remains the only publicly confirmed example. But as The New York Times reports today, cybersecurity experts are increasingly vocal that this was not an anomaly. It was a preview.

Both Anthropic and OpenAI are preparing to release more powerful AI systems in the coming weeks. These models can identify security vulnerabilities in computer systems far faster than any human, vastly raising the stakes in a decades-long war between attackers and defenders. The problem is obvious: the same technology that helps security teams patch holes also helps hackers find them.

This is the cybersecurity paradox nobody in the AI industry wants to talk about honestly. Every capability improvement is simultaneously a threat improvement. When your model gets better at writing code, it also gets better at writing exploits. When it gets better at understanding systems, it gets better at breaking them. There is no version of more powerful AI that is not also more powerful for attackers.

The timing of this story is remarkable. It lands alongside Anthropic warning governments that its next model could enable "large-scale cyberattacks," OpenAI burning $85 billion a year racing to build more powerful systems, and a New Yorker investigation revealing that OpenAI fired its CEO over safety concerns and then rehired him five days later under investor pressure.

The defense side is deploying AI too. Security companies are building AI agents that can monitor networks, detect anomalies, and respond to threats faster than human teams. But there is a fundamental asymmetry: attackers need to find one hole. Defenders need to find all of them.

We already reported that AI has made hacking nearly free. This confirms what that story predicted: the weaponization of AI for cyber operations is accelerating faster than the industry is willing to acknowledge publicly. The companies building these tools are also the ones most worried about how they will be used. That should tell you everything. First reported by The New York Times.