Catch Raises $5 Million to Build an AI Executive Assistant With Built-In Limits

Key points

• Catch, an AI startup, raised $5 million in funding to develop an AI-powered executive assistant.
• The product is built with access controls, meaning built-in rules that decide which company files, systems, and data the AI is allowed to use.
• The funding round was reported by SecurityWeek.
• Catch positions its product as a safer alternative to general-purpose AI tools that organisations may already be allowing staff to use without restrictions.

An AI startup called Catch has raised $5 million to build what it describes as an executive assistant powered by artificial intelligence, meaning software that can handle scheduling, research, and coordination tasks a human assistant would normally manage. The headline figure is modest by Silicon Valley standards. The pitch, though, is aimed squarely at a concern that keeps security teams awake: giving employees a powerful AI tool and losing track of what it can see.

Who is Catch, and what does the product actually do?

Catch is a new company. Details on its founding team and exact product stage are limited in early reporting. What the company says publicly is that its AI assistant is designed to behave like a trusted senior executive assistant, handling the kind of high-level coordination work that sits close to sensitive business information.

The key selling point is guardrails. In plain terms, that means the software has hard limits built in from the start, deciding which documents, inboxes, calendars, and internal systems it is permitted to access. It cannot roam freely through a company's data the way a misconfigured general-purpose AI tool might.

Why does this matter for ordinary workers and their organisations?

Most employees already use AI tools at work, often without their employer's security team knowing. That habit creates real risk. An AI assistant with no access limits can, in practice, read emails it was never meant to see, pull data from systems it has no business touching, or pass information to cloud servers outside the company's control.

Catch's argument is that access controls should be baked into the product rather than bolted on later. That is a reasonable position. Whether the implementation lives up to the claim is something security teams evaluating the product will want to test independently.

Should businesses be cautious before buying in?

Yes, at least until more detail is available. A $5 million raise is a seed-stage sum. The company is early. Claims about guardrails are only as good as the technical enforcement behind them, and right now there is limited independent verification of how Catch's controls actually work in a live corporate environment.

Organisations curious about the product should ask hard questions: What data does the AI send to Catch's own servers? Who audits the access rules? What happens if those rules are misconfigured?

For employees, the broader lesson holds regardless of which AI tool their employer chooses. Any software that connects to your work calendar, email, or internal files is a potential door into sensitive information. Knowing what your company's AI tools are permitted to access, and who decided that, is worth asking about.