“Something has gone completely wrong.”

Alex Karp, CNBC Squawk Box, July 1, 2026

The question your board is now asking.

That was Palantir's CEO on CNBC this week, describing how enterprises buy AI: paying for tokens, seeing little return, and watching their IP flow into models they don't control.

The clip traveled because the fear is real. Agents act with their connectors' permissions, at machine speed, and most companies keep no record of what they touched. When your board asks "what is our data doing inside these models?", the honest answer today is usually "we don't know."

Sovereignty is one route to control.

Sovereignty

Lock AI down and watch teams route around you — the tools reappear on personal accounts you can't see.

The gateway

If you run a fintech, you won't build a private model stack, and you don't need to. The data leaves through the connection between your agents and your systems. Control the connection and you control the exposure – while your teams keep using the models that make them fast.

What Metomic AI Gateway does.

Your agents – Claude, Cursor, and more – connect through Metomic to the systems they work with, with Slack and Google Workspace preconfigured and any remote MCP (Model Context Protocol) server supported.

Your agents

Claude Cursor

IBAN · PII

AI Gateway · in the path

See it

Govern it

Prove it

clean

Your tools & systems

Slack Google Workspace Remote MCP servers

See it.

From day one – before you write a single policy – you see every agent request, who and which agent made it, and what data it touched, plus the shadow AI your people are using in the browser. It all feeds into your SIEM.

Govern it.

Step in as it happens: coach, allow, block, or hold for human approval – with AI Judge handling the content rigid rules miss, and approved-AI rules setting which agents and tools get near your data.