AI agents are making cyber conflict harder to predict, says FT

In short: More cyber attacks are using AI agents, and that can make it harder for both sides to understand what is happening and how to respond.

What's going on

A Financial Times opinion column argues that cyber conflict is changing as attackers start using “AI agents”. An AI agent is software that can carry out a goal with less step by step human control, like giving an assistant a task and letting them decide how to do it.

The piece points to recent examples. The Australian government said AI agents breached parts of its healthcare systems. Anthropic, an AI company, also reported that Russia linked cyber groups were using AI for tasks like scouting targets, breaking in, and stealing data, at lower cost.

The writer says these attacks are not fully hands off yet, but the direction is clear. The more control that is handed to machines, the more difficult it becomes to tell what an attacker meant to do versus what the software did on its own. It can also be unclear whether an action is simple spying, preparation for sabotage, or just an automated system exploring whatever doors it finds open (like a burglar checking every window, even ones they did not plan to).

What to watch

The column warns that older ideas about deception, including Russia’s “maskirovka” tradition, may work differently when AI creates uncertainty for both sides. In a fast moving crisis, a defender might treat an AI agent’s unexpected move as escalation, and that misunderstanding could drive a bigger conflict. Expect more debate about limiting autonomy in offensive cyber tools and how to hold humans accountable when machines take more actions on their own.

Source: Financial Times