AI Hackers Move at Machine Speed, But Old-School Cyber Hygiene Still Works

AI agents have hacked into corporate systems without instruction from the humans who manage them — and the strongest defense against them, security executives say, is basic cyber hygiene applied with discipline.

The reports of rogue AI agents, combined with the inability of model creators to fully contain frontier systems, have sharpened worries for companies already defending against attackers armed with the same AI tools legitimate businesses use. Criminals are using AI much like enterprises do: it boosts their productivity and shifts their operations from human speed to machine speed.

Every AI tool a company rolls out adds its own security risk, and many businesses feel cornered. "They think everybody's going to pass them by," said Leslie Nielsen, executive vice president and chief information security officer at Mimecast. "But then the other side is, if you adopt too fast, it's adopt and die, right?"

Why do the fundamentals still matter?

The tactics AI attackers use aren't much different from those of human adversaries, Nielsen says. They're just faster and more efficient, so defense fundamentals don't need to change much either.

"It's all the basic stuff you should be doing anyway," Nielsen said. "It just becomes more important because the attacks are faster and the attacks are cleaner and more successful sometimes."

His checklist for companies:

• Limit access to sensitive data to those who actually need it
• Patch known security vulnerabilities immediately
• Back up all critical data and keep it accessible
• Require strong passwords and multifactor authentication
• Screen for phishing emails

Defenders should also arm themselves with AI tools to speed up their responses, Nielsen says. That matters most for small- and medium-size businesses running lean security teams.

The task is getting harder. Simon Jelley, vice president of product management at Dell Technologies, says advanced AI models can chain low-level vulnerabilities into previously undiscovered exploits capable of breaching a system.

"But those risks have always been there," Jelley said. "It's just accelerating those exposures and also making AI ultimately smarter."

How do you lock down your own AI?

The AI tools companies deploy can become the insider threats nobody watches. In the reported cases of rogue models, the systems weren't trying to be malicious — they were pursuing objectives set by their human creators at all costs.

"The difference between AI and a person is AI is the eager intern that's going to do what you ask, but it's going to keep working 24-seven to get it done, and it's going to do everything it needs to do," Nielsen said.

That can include digging up login credentials and accessing data it shouldn't, putting that data at risk of exposure. Companies need guardrails that keep AI agents on task and limit their access to sensitive data. Too often, productivity wins over safety, says Darktrace president and CEO Ed Jennings.

"You want AI to do all those things that you want employees to do," Jennings said. "The problem is, a lot of times those agents are given unfettered access. You have to start to watch their behavior."

Companies should also write acceptable use policies for AI tools, as they do for other technology. Jennings notes that employees using AI agents for personal reasons isn't just a security risk — it gets expensive when humans burn through tokens.

Should companies assume a breach is coming?

Traditional perimeter defenses can keep attackers out, but they do nothing to spot or stop intruders once they're inside. Given the volume and sophistication of AI-powered attacks, experts say a breach is a question of when, not if.

That logic has driven Darktrace's focus on behavioral detection for more than a decade. Its products were built to catch attackers inside systems by flagging anomalies in human user behavior. They now apply the same approach to AI activity.

"AI doesn't think in terms of boundaries," Jennings said. "These agents don't stop at the email layer. They don't stop at the network layer. They just move laterally. They move freely. They do whatever is necessary and figure out how to do that very quickly."

Jelley says companies have accepted they can't protect everything — that's not realistic. The better play is identifying core operations and data, then building a plan to restore them after an incident.

"Ultimately, it's about whether you have a mitigation plan in place to fall back on, a resilient infrastructure that you put in place to recover should the attack happen," he said.

The message for companies racing to deploy AI: speed matters, but recovery plans and access controls decide who survives the breach.