AI agents are going rogue, and the lawsuits are going to get messy

It's a new fact of life that AI agents are bypassing their sandboxes to get into places they aren't supposed to. We've learned to stomach (or largely ignore) cyberattacks as part of the cost of doing business, or really, existing, while being connected to the internet.

But AI agents present a new dilemma. Rather than bad actors probing vulnerabilities, wielding whatever tools they can, the agents are the sophisticated products of some of the most influential and powerful tech companies in the world.

"Agents going rogue is the biggest risk today. Imagine an agent on your corporate network hacked or hijacked or it goes rogue," Zscaler (ZS) founder and CEO Jay Chaudhry said on Yahoo Finance's Sozzi Unleashed. "Don't trust AI agents."

Every week, it seems, we're seeing sheepish disclosures by OpenAI (OPAI.PVT) and co. that its agents "accidentally" hacked something it shouldn't have.

A society has to grapple with some element of crime. But what if those criminal acts are in part perpetuated by the supposedly upstanding corporations powering the S&P 500? While defense against rogue agents may feel futile, this is America, and we can always go on offense and throw the book.

But rogue AI, liability, and lawsuits pose a set of thorny questions that the legal system has yet to confront.