Four of five enterprises that secured AI agent identities still can't contain one that goes rogue
Visa's president of technology, Rajat Taneja, walked the VB Transform 2026 audience through aiming Anthropic's Mythos at Visa's own payment network . The model stitched minor weaknesses into working exploit chains, and Visa open-sourced the harness that governed the hunt. That's what it looks like when an enterprise has the engineering depth to act on what it finds. Most don't get there. Just…
More than half of enterprises have experienced an agentic security incident or near-miss, despite 65% enforcing agent permissions at runtime. Isolating high-risk agents is rare, with only 8% pairing enforcement with isolation. Research indicates that enterprises rely on provider-native controls, further widening the containment gap.
Despite rating their security tools higher, those who experienced breaches gave them higher satisfaction scores. The fast improvement in isolated agent adoption has yet to translate into higher satisfaction for enterprises.
Written by urgent.news from VentureBeat's reporting — not their text. Machine-written — may contain errors; check the original before relying on it.
