A tip from Brenda Christensen at RSA Conference 2026 pointed me toward Airrived, a Dublin, California startup that announced a new product called AetherClaw at the show. Here is what Airrived actually does, in plain language.

Start with the problem

Artificial intelligence agents are no longer theoretical. Companies are deploying systems that can take real actions: modifying access permissions, pulling sensitive records, triggering workflows, sending communications. The tools that made it easy to build these agents quickly were designed for exactly that—speed and flexibility. What they were not designed for was telling you what an agent was allowed to do, creating a record you could show an auditor, or giving a human a way to intervene before an action was taken.

For a proof of concept, that is fine. For a hospital, a bank, or a utility company, it is not.

What Airrived built

Airrived sits between an artificial intelligence agent and your enterprise systems. Think of it as three things at once: a rulebook that defines what an agent is permitted to do, a checkpoint that enforces those rules before any action executes, and a recorder that logs every decision in a format an auditor can read later.

They call this layer the Agentic Operating System. AetherClaw is the security-specific capability built on top of it, targeting four domains where the stakes of a misbehaving agent are highest: identity and access management, vulnerability management, security operations, and governance, risk, and compliance.

The company is not building the agents themselves. They are building the governance wrapper that makes agents safe enough to run in regulated environments. Their stated customers are enterprises in financial services, healthcare, and critical infrastructure.

Why this showed up at RSA 2026

Airrived was not the only company at RSA Conference 2026 addressing agent governance. Cisco, Microsoft, CrowdStrike, and SentinelOne all announced capabilities in the same space. The naming coincidence was striking: Cisco called its offering DefenseClaw, Nvidia has OpenClaw (with NemoClaw adding enterprise security controls), and now Airrived has AetherClaw. The "Claw" suffix has apparently become the unofficial badge of the agentic security movement.

What distinguishes Airrived from those larger players is the founding premise. The enterprise vendors are adding governance layers onto platforms that were built for other purposes. Airrived built governance-first from the start, rather than retrofitting it onto an existing product. That is a legitimate architectural claim. Whether it holds up at enterprise scale against players with existing security relationships is the question any buyer should be asking.