Former Cisco artificial intelligence security researchers have launched a new company to tackle a problem that barely existed a year ago: securing the autonomous AI agents enterprises are handing the keys to their most critical systems.

Tenet Security Inc. today formally launched a platform designed to stop malicious AI agent behavior before it reaches production systems. Its core technology, called Agent-Side Simulation, simulates an agent’s likely next moves before they hit live infrastructure.

If a path looks risky, Tenet intervenes before damage occurs and ships a trace explaining why the action was blocked. Most security tools raise an alarm after something suspicious has already happened. Tenet tries to catch the problem first, before the agent actually does anything.

Companies are giving AI agents real power to act on their own. The agents run code, pull from databases and make changes to live systems. The catch is that once an agent is loose in the environment, security teams mostly cannot tell what it’s doing. Tenet says organizations often have as many as five times more AI agents operating than their security teams realize.

“AI agents may be the biggest productivity unlock enterprises have seen in decades, which is why organizations are moving so quickly to deploy them,” said co-founder and Chief Executive Barak Sternberg. “But we’re also entering a world where autonomous agents are interacting with systems, data and other agents in ways most security tools were never designed to understand.”

Sternberg founded Tenet with Nevo Poran. Both are offensive security researchers who worked on Cisco’s AI Defense and studied how attackers go after autonomous systems. Earlier they ran Wild Pointer, a cybersecurity company that reached seven-figure annual revenue with Fortune 500 customers on its books. Sternberg and Poran have each spoken at DEF CON and Black Hat.

The launch follows research from the company’s Tenet Threat Labs into what it calls “agentjacking,” a class of attack that manipulates agents into executing attacker-controlled actions.

The attack hides a malicious instruction inside something the agent reads, maybe an email, a log entry, or a record in a database. When the agent gets to it, the instruction takes over and redirects what the agent does next. T

enet says it tested this across more than 100 enterprise environments and found thousands of organizations that could be hit. Traditional security tools missed it. The agents were doing what they were allowed to do, so nothing tripped an alarm.

Early deployments point to the scale of the problem. Tenet says one legal-sector enterprise with $1 billion in annual recurring revenue grew from two agent deployments to more than 20 over six months while using the platform, which blocked more than 10 attempted attacks, including a critical cross-site scripting attempt. At a separate Fortune 1000 customer, Tenet flagged a runaway agent burning tens of thousands of dollars in token consumption over a single weekend.

“We’re increasingly seeing AI agents become part of the attack path itself,” Poran said. “The only place left to catch these threats is at runtime, in the moment an agent decides to act.”

The launch was backed by $6 million in seed funding to support product development, expansion of Tenet Threat Labs and growth of the company’s North American go-to-market operations. The Westly Group, an early backer of SentinelOne Holdings Inc., led the round with participation from MizMaa Ventures Ltd. Tenet’s advisers include David Schwed, former chief information security officer at Robinhood Markets Inc., and Rick Scott, former CISO at The Bank of New York Mellon Corp.

Photo: Tenet Security