We're handing agents the keys to production systems and hoping logs are good enough. They're not.

The common thread in every MCP security incident: no visibility into what the agent was doing until after the damage was done.

You're building systems to help real users but you can't observe real usage. You can't build feedback loops because the feedback itself contains the data you're not allowed to look at.

Hundreds of millions of patient records sitting across hospital systems. The best training dataset in medicine, and no way to actually train on it.