Former Cisco artificial intelligence security researchers have launched Tenet Security Inc., a startup addressing a fast-spreading issue: controlling the actions of autonomous AI agents before they cause harm.
Its platform uses Agent-Side Simulation to predict an agent’s next moves in production systems. If a path appears risky, Tenet blocks it and provides a trace explaining the decision. Unlike most security tools that respond after an incident, the startup prevents threats before they occur.
Businesses are increasingly granting AI agents access to critical systems, letting them execute code, query databases, and modify live environments. The problem is that once deployed, these agents often operate beyond the oversight of security teams. Tenet states many organizations underestimate their exposure, with some hosting far more agents than they realize.
Co-founder and CEO Barak Sternberg noted that AI agents may be the largest productivity gain enterprises have seen in decades. He added that the world is entering an era where autonomous agents interact with systems, data, and other agents in ways most security tools were not built to handle.
Related: Arcade gets 60 million dollar boost
Sternberg and co-founder Nevo Poran previously worked on Cisco’s AI Defense team, researching attacks on autonomous systems. Earlier, they ran Wild Pointer, a cybersecurity firm serving Fortune 500 clients with seven-figure annual revenue. Both have spoken at DEF CON and Black Hat.
Tenet Threat Labs, the company’s research division, discovered a new attack method called “agentjacking,” where attackers hide malicious instructions in data an agent processes—such as an email, log, or database record. Once read, the hidden command changes the agent’s behavior.
Testing across over 100 enterprise environments revealed thousands of vulnerable organizations. Traditional security tools missed these attacks because the agents were performing allowed actions, triggering no alerts.
Early customers have already benefited. A legal-sector enterprise with high annual revenue expanded from two to over 20 agent deployments in six months using Tenet’s platform. The system blocked multiple attacks, including a serious cross-site scripting attempt. At a Fortune 1000 client, Tenet identified an agent wasting significant funds on token usage over a weekend.
Related: Helion startup valued at 15 billion dollars
Poran stated that AI agents are increasingly part of the attack path. He explained that the only effective place to stop these threats is at runtime, as an agent decides to act.
Tenet raised $6 million in seed funding led by The Westly Group, an early backer of SentinelOne. MizMaa Ventures also joined the round. Advisers include David Schwed, former CISO at Robinhood, and Rick Scott, former CISO at Bank of New York Mellon.
The funds will support product development, expansion of Tenet Threat Labs, and growth in North America.
