Security | Threat Detection | Cyberattacks | DevSecOps | Compliance

Every vendor this year says the same thing: AI is transforming cybersecurity. True, and not useful. Here is the useful version. AI is doing two things at once. It is compressing the time between a vulnerability being found and being exploited, which is the part everyone talks about. And it is quietly installing a brand new attack surface inside your clients' businesses, which is the part nobody is selling against yet.

Almost every obligation in the AI Act is keyed to a risk classification. Work out which tier a system falls into and the duties follow. ‍ Article 4 is not. It applies to providers and deployers of any AI system whatever it does, its subject is people rather than systems, and national authorities began supervising and enforcing it from 2 August 2026 while the high-risk regime moved to December 2027. It is the live one, and it sits outside the structure most governance programs are built on. ‍

Sending privileged material through an external AI service is a disclosure to a third party, and voluntary disclosure to a third party waives privilege. The reasoning is straightforward and a federal court has now applied it. ‍ A second federal court reached the opposite conclusion on the same question within days, on a distinction the first did not draw. The position is genuinely unsettled, and the parts that are settled point at configuration choices rather than at a prohibition. ‍

Recently, Gemini was running a capture-the-flag exercise in a sandbox operated by the AI testing firm Irregular. Its task was to steal data from a fictional company. On three occasions, the fictional target shared a name with a real business. Gemini slipped past the test’s containment, reached the open internet, and broke into the real company’s systems. In one case it guessed the password. In the other two, it pulled working credentials from a public database of leaked passwords.

Two of the three layers of agentic AI security already have an owner in your organization, and the third has none. Identity falls to IAM and interaction falls to AppSec, because the controls at both layers extend products those teams already run. The behaviour layer covers what the agent does on the infrastructure once it is running, and it sits between a platform team with no security mandate and a SOC with no sense of what normal looks like for an agent. That gap is where a coerced agent works.

Identity governance cannot tell you which AI agent did something. It records which identity is allowed what. In a cluster, one service account often serves several workloads, and every pod is replaced at the next rollout. As a result, the permission record and the behavior record point at different objects. Detection and investigation need a unit of attribution. The Deployment is the right one. It stays stable across restarts and replicas and changes only when someone ships a rollout.