**Our Take: The Containment Gap Is the Story**
The data from this wave of VentureBeat Pulse Research is clear: enterprises have built the first two layers of agentic security and largely skipped the third. Watching and permissioning are now table stakes, 65% enforce scoped permissions at runtime and 56% monitor agent activity. But only 18% isolate their highest-risk agents, and just 8% pair enforcement with isolation. That ordering is backward from a defense-in-depth standpoint. Observation tells you what happened. Enforcement tries to stop it. Isolation is what limits the damage when enforcement fails, and the near-misses in this report prove it will fail. A fleet that is watched and permissioned but not boxed in is precisely the configuration where a single control failure propagates across systems.
The uncomfortable pairing is confidence with exposure. Satisfaction is at a series high of 4.29 out of 5, yet 63% rate the contest against AI-armed attackers as even or worse, and 30% say attackers are ahead outright. More telling, 74% plan to replace tooling they just rated highly. That is not satisfaction; that is comfort with what is easy. The stack remains 92% provider-native by primary layer, and the specialists built for exactly these gaps sit in single digits. Identity, the structural weakness beneath most incidents, is still the least present in purchase plans. Credential sharing persists across nearly two-thirds of fleets, and only 29% describe scoped identities with no sharing anywhere. The organizations with the widest blast radius are not meaningfully more likely to have built the control that bounds it.
The pattern is not about tooling. It is about intent. Enterprises are more determined than ever to change a stack they are more satisfied with than ever, and the gap between those two facts is where the next incident lives. The consideration set has broadened, but not in the direction the incident data points. Agent-identity products are in just 10% of consideration sets, and isolation remains the least-built control even among those who have already been hit. Experience remains the strongest predictor of urgency, but urgency has not yet translated into action where it matters most.
If you are running agents in production, the question is not whether you have a near-miss ahead of you. It is whether the next one stays a near-miss. The control that answers that question is the one most of you have not built yet. The data is in. The gap is yours to close.