Tenet Security, an AI security startup founded by former Cisco cybersecurity researchers Barak Stern...

Tenet Security, an AI security startup founded by former Cisco cybersecurity researchers Barak Sternberg and Nevo Poran, has emerged from stealth with $6 million in seed funding. The round — led by The Westly Group and MizMaa Venture with participation from Dazz founder Tomer Schwartz and former Coralogix CEO Lior Tal — backs a platform that monitors, pre-analyzes, and blocks risky actions by autonomous AI agents before execution. The company also published research on a novel attack vector it calls "Agentjacking," where adversaries can manipulate AI agents into carrying out unauthorized actions on their behalf.

Why it matters: Tenet's launch lands at the precise inflection point where enterprise AI agents — capable of accessing corporate databases, codebases, and internal systems with limited human oversight — are becoming the norm. As AI deployment shifts from experimental pilots to production workloads, the security paradigm is also shifting: the threat surface is no longer the model itself but the autonomous actions the agent can take on behalf of a user. Tenet's "pre-execution guardrail" approach, combined with its agentjacking research, directly addresses the emerging governance gap that security leaders have flagged as a top concern for 2025-2026 agentic deployments.

Grounding the analysis: The $6M seed is not a structural capital event — it falls well below the $500M threshold for a cross.§D tag. But it does slot neatly into the AI Agent security sub-segment (Segment 02) and updates the player map for agent-infrastructure guardrails. Tenet's research on agent-level manipulation also contributes to the growing corpus of attack-surface documentation (#5.7 pattern — "new attack surface before the market matures"), a recurring pattern where early-stage startups define a threat taxonomy that later incumbents acquire or rebuild. The company's customer references — including cost-overspend from rogue agents — align with the "context-engineering moat" thesis: the best agent security tools are those that watch what the agent actually does, not what it says it will do.