Report says AI agents made it harder to spot government site hacks

In short: A new report says attackers used AI agents during government website breaches in ways that helped hide signs of the hacking.

What happened

Researchers at Asymmetric Security said they found more evidence that attackers are using AI agents when breaking into government websites. An AI agent is software that can carry out steps on its own after being given a goal, like a helper that follows a checklist.

According to the findings, the AI agents obscured signs of the break-ins. In practice, that can mean the normal clues security teams look for were harder to spot, or looked less suspicious than usual. Think of it like someone entering a building while carefully cleaning up footprints as they walk.

The Financial Times reported on these findings as part of a wider look at how AI tools are changing hacking tactics. While the source article focuses on OpenAI’s agents, the new context from Asymmetric Security points to a broader pattern, where AI-powered tools can help attackers move faster and blend in.

Why it matters

Government websites can hold sensitive information and provide public services. If attacks become harder to detect, it can take longer for defenders to realize something is wrong, and longer to stop it. For the public, that can raise risks like service outages, stolen data, or a loss of trust in online government services.

Source: Financial Times