Introducing AI Proxy: See and Secure the Shadow AI Already on Your Network
By Krista Lyons
We're building a way to see, govern, and secure how your team actually uses AI — before it becomes a problem you find out about the hard way. Here's what AI Proxy does, and how to get on the Beta list.
|
The short answer: AI Proxy is a new CloudConnexa capability that shows admins which employees are sending what to which AI services — ChatGPT, Claude, and others — and lets them set policy, block risky prompts, cap usage, and generate compliance evidence, without installing a separate agent. It's moving from internal testing into Beta. Take the two-minute AI Proxy Beta interest survey to help shape it and get notified the moment it opens. |
Shadow AI is already inside your network
Somewhere in your company right now, someone is pasting a customer contract into ChatGPT to summarize it. Someone else is feeding source code into an AI coding assistant. A third person uploaded a spreadsheet of prospect data into a free AI tool nobody in IT approved, reviewed, or has even heard of.
It's not hypothetical, and it's not rare. In one survey, 48% of working adults admitted to entering private company information into generative AI tools — and that share is almost certainly higher today (see Generative AI Security Trends). Most of that traffic rides over an ordinary HTTPS connection, in a browser tab your team already trusts, so it never shows up in a log your security team is watching. You can't govern what you can't see — and right now, most SMB and mid-market IT teams can't see any of it.
A compliance clock started running this year
Meanwhile, the regulatory pressure isn't theoretical either. The EU AI Act phases in major obligations for general-purpose and high-risk AI systems across 2026 and 2027, Colorado's AI Act took effect February 1, 2026, and frameworks like ISO/IEC 42001 and the NIST AI RMF are becoming the reference points procurement and audit teams reach for. Sooner or later, a customer's security questionnaire or your cyber insurer is going to ask a simple question:
What AI is your team using, and what have they sent it?
Today, most companies don't have an answer. Soon, not having one won't just be a blind spot — it'll be a gap someone else notices first.
What AI Proxy actually does
AI Proxy is built to answer that question, in plain language:
- See what's happening. One view of which users are sending what to which AI service — no endpoint agents, no separate vendor portal to check.
- Decide who can use what. Label AI vendors and tools as sanctioned, unsanctioned, or monitor-only, by team or user group.
- Stop the wrong content from leaving. Block prompts that contain content you've defined as off-limits — starting with a simple keyword list, growing to full PII, secrets, and source-code detection.
- Cap what AI usage costs. Set usage limits per device, so one runaway integration doesn't turn into a runaway bill.
- Prove it. Generate an audit trail and evidence you can hand to an auditor, a cyber insurer, or a customer's security review — instead of assembling one from memory after the fact.
Built into the connection you already have
AI Proxy runs on the CloudConnexa connection your team already uses to reach the internet securely — it isn't a new agent to deploy or a new console to log into. It uses the identity and User Groups you've already set up, shows up as a new section inside the CloudConnexa Admin Portal, and reports through Log Streaming like everything else in your tenant.
Why this, why now?
Every major SASE/SSE vendor added some form of AI security in 2025 — but bundled into six-figure enterprise contracts. If you're running an SMB or mid-market IT team, the honest answer you've probably gotten so far is "wait" or "buy the bigger platform to get the one feature you need." AI Proxy exists to close that gap at CloudConnexa's existing pricing, without asking you to move to a bigger plan. It's a natural extension of the zero-trust network access model CloudConnexa already applies to every other connection in your network — just pointed at AI traffic instead.
Get on the Beta list
AI Proxy is moving out of internal testing and toward Beta. Before it opens, we want to hear from the people who'll actually use it: take our two-minute AI Proxy Beta interest survey and tell us which AI vendors you need visibility into, what "sensitive" means for your data, and what would make you comfortable turning this on. Everyone who responds gets first notice when the Beta opens — and a direct line to shape what ships.
We want to hear from you.
Take our 2-minute survey and tell us which AI vendors you need visibility into.
Take the SurveyFAQ
Do I need to install anything for AI Proxy to work?
No. It's built into the CloudConnexa connection your team already uses — there's no separate agent or client to deploy on each device.
Will AI Proxy block my team from using AI tools altogether?
No. Nothing changes by default until an admin sets policy. You decide which AI services are sanctioned, monitored, or blocked, and you can start with visibility alone before you turn on any blocking.
What will it cost?
Planned pricing is flat, per seat, per month, with a baseline amount of AI usage included — optional top-up packs cover teams that go beyond it. Final pricing will be confirmed at Beta.
When does the Beta open, and how do I get access?
We're not committing to a public date yet. Take the interest survey above — respondents get first notice when the Beta opens and help shape which capabilities ship first.