Avanoo V3: Three Pillars to Take Back Control of Shadow AI

In 2026, the question is no longer whether AI has a place in the enterprise. It's where it's hiding. Language models used outside any framework, autonomous agents plugged into internal systems, MCP servers spun up by technical teams in a matter of hours: AI is moving faster than the policies meant to govern it.

Avanoo is announcing the launch of its V3 platform, built to close that gap. Three products make it up:

• Discover for mapping
• MCP Proxy for controlling agents
• AI DLP for protecting data at the prompt level

All the new features will be presented in detail during our launch webinar on September 8 at 11am (CET). Register now!

Pillar 1 — Avanoo Discover: mapping for the agentic era

Avanoo Discover is our flagship product: a browser extension deployed silently across your fleet, producing a complete map of the tools your employees actually use.

V3 significantly expands this scope.

Beyond standard web applications, Avanoo Discover now detects:

• AI agents active in the work environment, whether built into business tools or deployed autonomously.
• Running MCP servers — the standard protocol that lets agents interact with your internal systems (databases, APIs, business tools...).
• The authentication mechanisms actually used, application by application: SSO, SAML, MFA, OAuth.
• Risky behaviors tied to AI tools: reuse of corporate passwords on consumer AI tools, data copy-pasted or files uploaded to unapproved tools, and automatic separation between professional and personal AI accounts used from the same workstation.

The new AI Dashboard and built-in AI Chat flip the usual model on its head: information comes to teams in real time, in natural language, instead of teams having to dig through static tables.

Pillar 2 — Avanoo MCP Proxy: the control plane for AI agents

This is the most strategic addition in this release.

The Model Context Protocol (MCP) has become the de facto standard for letting AI agents interact with third-party systems. Its rapid adoption has created a major blind spot: traditional cybersecurity solutions were never designed to see it, let alone control it.

Avanoo MCP Proxy positions itself as a control plane sitting between AI agents and your information system. In practice, it lets you:

• See which agents are calling which MCP servers, and with what level of access.
• Filter requests against the security policies your organization defines.
• Audit every interaction in real time.

For organizations rolling out copilots and automated workflows at scale — often without formal governance over this infrastructure — this is a direct answer to a new, still largely underestimated risk.

Pillar 3 — Avanoo AI DLP: protection moves up to the prompt

The third building block is currently in development.

Avanoo AI DLP operates at a level traditional DLP never reaches: the prompt itself, before information ever leaves the organization.

The platform analyzes content submitted to browser-based AI tools to detect sensitive data: HR records, financial information, proprietary source code, contract numbers, and more.

Depending on the policies defined, it can block, alert, or log the activity.

Avanoo is building this module in partnership with the French cybersecurity ecosystem, as part of ANSSI's AMI CYBERAI_2026 program.

A European answer

The tools driving AI adoption in the enterprise are, for the most part, built outside Europe. That dependency isn't neutral: it weighs on data control, on the ability to meet European regulatory obligations, and on the strategic autonomy of the organizations that rely on it.

Avanoo, a French company, is taking the opposite bet with its MCP Proxy and AI DLP modules. Both products — which touch the most sensitive flows (AI agents, prompts) — deploy on-premise, with no data routed through a non-European provider: an architecture built for organizations that can't afford to depend on infrastructure outside their control.

For regulated organizations (banking, insurance, public sector, healthcare), this is an argument that matters more and more in security architecture decisions: a concrete answer to growing compliance obligations (GDPR, NIS2, DORA, the AI Act), with on-premise deployment for the modules handling the most sensitive data.

Join us on September 8

We'll see you on September 8 at 11am (CET) to discover Avanoo V3 live.

On the agenda: a look at the battle currently playing out among AI vendors and what it means for your IT environment, a field perspective from Jérôme Quinten, CISO at Ponant, a live demo of the platform and its new features, and a session on the next frontier of AI security with Léonard Kéat, cybersecurity and AI ethics expert at Advens.

Seats are limited: register now.