CharliAI and Protegrity will sell banks and insurers one tool to limit what AI agents can see and do, aiming to get stalled projects past risk review.
CharliAI and Protegrity announced a partnership on Oct. 20, 2026 to help banks, capital markets firms and insurers put AI agents to work on sensitive data. An AI agent is software that carries out tasks on its own instead of just answering questions. The pitch is simple: get agents past the risk reviews that keep holding them back.
The two firms are combining their tools. Protegrity protects the sensitive data agents use, and the knowledge they could piece together from it. CharliAI controls what agents are allowed to do and keeps a detailed record of what they did.
Banks, capital markets firms and insurers can buy the combined product from CharliAI. It resells Protegrity's data and AI security tools alongside its own platform, called Ancaeus.
The announcement cites research from Enterprise Management Associates (EMA). In its 2026 study, 83% of IT and security leaders said security or compliance reviews have delayed AI from reaching production. Also, 69% already run agents that write to databases and trigger outside systems. Many still rely on controls built for people looking up data, not for agents acting on it.
Protegrity CEO Michael Howard framed the risk this way. The agent to worry about is not only one that goes rogue. It may be one that does exactly what it is allowed to do and works out things nobody meant it to know. He called every agent a new insider that never signed an NDA, a promise to keep secrets.
The companies say review comes down to two questions: what can the agent see, and what can it do? A one-time access check cannot answer them, because agents reason, hand work to other agents and act at machine speed. A dashboard that reports problems after the fact falls short too.
Their answer is to build protection into the agent's workflow rather than at the edge of the system. CharliAI CEO Kevin Collins put it this way: "capability is not clearance." An agent that can run a complex task does not automatically have the authority to do it unsupervised.
The announcement gives no price, no start date and no names of banks or insurers using the product. The performance claims come from the companies themselves. How well the combined tool works in a real review has not been shown in the material.
If your business uses AI that acts on its own, ask who sets its limits and who keeps the record of what it did.
OpenAI has about $11 billion in hand at a $1.25 trillion value, below its first ask. Microsoft and NVIDIA shareholders have reason to watch what comes next.
Read more: OpenAI raises $11 billion, but at a lower price than it wantedVisa tested 22 AI models on realistic shopping tasks. They passed just over half, and often bought the wrong thing anyway.
Read more: Visa tested 22 AI shoppers, and most still make costly mistakesDARPA wants AI to hunt for design flaws in stock, bond, energy and cloud markets, so enemies cannot use them to cause economic damage that reaches households.
Read more: Pentagon wants AI to find weak spots in stock and energy markets
