Reco launches Browser Guard for enterprise AI security

Reco has launched Browser Guard, a browser-based security product for enterprise AI use designed to address prompt injection, data leakage and unauthorised actions.

The launch reflects a shift in how employees use artificial intelligence tools at work, with much of that activity now taking place in web browsers rather than in software already monitored by security teams.

Browser Guard inspects prompts and agent activity in the browser, then matches that activity with information about the user, the data involved, the destination, the account type and the stated business purpose, according to Reco.

The product is aimed at organisations trying to manage the spread of AI assistants, copilots, coding agents and automated workflows used by staff in day-to-day tasks. These interactions can create risks when security teams lack visibility into who is using a tool, what information is being entered and where the resulting actions are directed.

Browser focus

Reco is targeting a part of the market where AI use often begins outside approved channels. Browser Guard can identify AI tools, browser extensions and personal accounts used by employees beyond sanctioned systems, giving security teams visibility into whether business data is being shared.

That focus on browser activity addresses what security specialists often describe as shadow AI, where workers adopt consumer or personal AI services without formal approval. In those cases, the concern is not only the tool itself but also where company data is sent and how far automated agents can act on a user's behalf.

Agents create a broader set of risks than a single prompt because they can retrieve context, call tools and carry out actions, Reco said. Browser Guard is intended to assess the identity behind an agent's action, the data it can access and whether the behaviour aligns with company policy before allowing, guiding or blocking the activity.

Policy controls

The product adds a layer of context to each interaction by identifying the user, the data, the destination, the account type and the business purpose behind it. That information is intended to help security teams make policy decisions as the activity takes place rather than after the event.

For companies adopting AI tools across business functions, the challenge has grown beyond maintaining lists of approved software. Security teams are increasingly being asked to judge the behaviour of AI systems embedded in browsers, messaging tools and coding environments, especially when those systems can trigger external actions or move data between services.

"AI security is not just about knowing which tools employees use," said Ofer Klein, Co-Founder and Chief Executive Officer of Reco.

"Security teams need to know what data enters each interaction, what an agent can reach and whether its next action should be allowed. Browser Guard enforces policy decisions in the browser, where AI use is actually happening, with the context needed to stop risky activity before it creates exposure," Klein said.

Broader platform

Reco positions the new product within a wider security platform focused on what it calls the agent ecosystem, where applications, identities, permissions and workflows intersect. Its platform includes more than 280 app integrations and 1,000 detection controls, according to the company.

That broader approach suggests Reco is seeking to tie browser-based AI oversight to existing efforts to map third-party software, user identities and access rights across large organisations. As companies test more autonomous tools, the ability to connect browser activity to identity and policy data is likely to become a more contested part of the security market.

Browser Guard is now generally available.