Exaforce has launched ExaU, a service that lets security staff create teams of AI agents for investigations and other security work. Its notable guardrail: agents can investigate on their own, but actions affecting employees, systems or communications require approval.
Watch Desk analysis
What happened
SiliconANGLE reports that staff can create agents by describing a job in plain language, or start from templates. Agents are designed to retain decisions, exceptions and past instructions, and can be shared across a security team. Exaforce’s example has a finding trigger a threat hunt, an activity audit and outreach to an employee, with one assessment bringing the results together.
The service runs on customers’ data within the Exaforce platform and is available now. The company says its audit log records each step for later review. ExaU follows Exaforce’s September launch of a kill switch for AI agents that go rogue.
Why it matters
This is a move from agents handling isolated security tasks towards coordinating several strands of work around one finding. That could help teams tackle backlogs, while approval requirements and action records offer people a chance to oversee consequential steps. The announcement describes the product’s design, not independent evidence of how well it performs in practice.
Our read
The useful question is whether a team can delegate investigations without losing track of what its agents are doing. ExaU’s approval rule and audit log are sensible ingredients; the real test is whether they make complex work easier to review rather than simply adding another place to check.
What to watch
- How customers use shared agents across security teams.
- Whether Exaforce publishes evidence of investigation time saved or results improved.
- How approval rules and audit records work in real deployments. Activity teaser: Exaforce’s new ExaU service lets security teams create and share AI agents for investigations. Agents can investigate independently, but actions affecting people, systems or communications need approval first. That is a concrete attempt to pair delegation with oversight, rather than trusting an agent to know when it has gone too far. Our full story looks at what the service can do and what still needs proving.
Discussion spark: Should security teams allow AI agents to investigate independently when consequential actions still need human approval, or should people review every step?
Sources and evidence
- Exaforce's ExaU lets security staff spin up a personal team of AI agents – SiliconANGLE (16 November 2026, 12:00 UTC)
Watch Desk is operated by WittyWires as an independent cross-cutting AI news tracker. It does not speak for the organisations or people it covers.