Browser security is now the hidden control plane for identity risk

At a glance

What this is: This is an analysis of how browser activity has become the main place where identity, GenAI, extensions, and data leakage intersect, with traditional controls missing much of the risk.

Why it matters: It matters because IAM, DLP, and session governance increasingly need browser-level visibility to manage NHI, human, and AI-driven access paths that now sit outside conventional control points.

Context

The browser has become the operating surface for identity activity, SaaS access, GenAI prompts, extensions, and session-based data movement. That changes the governance problem because many security programmes still treat the browser as a presentation layer rather than a control plane where identity, data, and session state are actively reused and exposed.

Traditional identity controls were built around authentication and central policy enforcement, but they do not fully observe what users do after login. When prompts, copy/paste, cookies, and unmanaged extensions carry sensitive data or active session context, the control boundary shifts from the identity provider into the browser itself.

Key questions

Q: How should security teams govern browser-based policy enforcement for identity and data risk?

A: Start by classifying which decisions belong at session time rather than at login, then assign ownership across identity, endpoint, data, and SOC teams. The browser should enforce only policy that is clearly defined, logged, and reversible. That prevents the browser from becoming an opaque shadow control plane.

Q: Why do browser sessions create exfiltration risk even when SSO and MFA are in place?

A: Because SSO and MFA protect entry, not everything that happens inside an authenticated browser session. Once cookies, cached tokens, and SaaS tabs are active, data can move through copy/paste, prompts, and extensions without a new login event. The risk is reuse of trusted state, not just credential theft.

Q: What breaks when browser extensions are allowed broad access to cookies and tabs?

A: Browser extensions can act like unmanaged software dependencies with access to identity-bearing session data. When permissions are broad, a malicious or compromised extension can read page content, harvest tokens, or exfiltrate data silently. That breaks assumptions about the browser being a passive client and turns it into an active risk layer.

Q: What happens when employees upload sensitive data into GenAI tools from the browser?

A: Sensitive information can leave the organisation in ways traditional perimeter controls do not reliably stop. Once users can paste, upload, or share material into external GenAI services, confidential data may be exposed outside approved governance. The safest response is to apply browser-level restrictions and masking controls before information reaches the tool.

Technical breakdown

Why browser sessions have become the real identity control plane

A browser session now holds more security-relevant state than the login event does. Cookies, tokens, autofill data, SaaS tabs, and cached session context let users move between applications without re-authenticating, which means identity risk accumulates after SSO and MFA complete. That creates a governance gap for browser-native telemetry, because the security event is no longer the authentication itself but the reuse of authenticated state across tabs, prompts, and third-party services. Practical implication: instrument the session, not just the login.

Practical implication: extend identity monitoring into active browser sessions where token reuse and account crossover actually occur.

How GenAI prompts and copy/paste become exfiltration paths

The article shows that prompt-driven workflows are now a data movement channel, not just a productivity feature. When users paste corporate content into GenAI tools, the browser becomes the transport layer for sensitive text, files, and context that traditional DLP often misses because it is not packaged as an attachment or email. This is especially important when personal accounts, unmanaged browsers, and chat-style interfaces blur the line between sanctioned and unsanctioned use. Practical implication: build controls around user actions inside the browser, not just file endpoints.

Practical implication: apply browser-native DLP to prompt input, copy/paste, drag-and-drop, and account context.

Why browser extensions behave like an unmanaged software supply chain

Extensions run with broad access to tabs, cookies, and page content, which makes them more like embedded software dependencies than harmless add-ons. The article’s point is that publishers, update cadence, sideloading, and permissions create a supply-chain style risk inside the browser itself. That is why permission review alone is weak: the real control problem is provenance, update trust, and runtime access to identity-bearing session material. Practical implication: govern extensions as third-party software with continuous inventory and trust scoring.

Practical implication: treat extension provenance and update behavior as part of your software supply-chain governance.

Threat narrative

Attacker objective: The attacker aims to harvest browser-held session material or sensitive content to impersonate users and exfiltrate data without triggering conventional controls.

1. Entry occurs through normal browser use, where employees authenticate into SaaS apps, open GenAI tools, or install extensions that gain access to session state and page content.
2. Credential and session material are then exposed through cookies, tokens, prompts, or extension permissions that let attackers or unsanctioned tools reuse authenticated browser context.
3. Escalation follows when that browser context is used to impersonate users, move across SaaS apps, or silently transmit sensitive data outside governed channels.
4. Impact is data exfiltration, session theft, and identity misuse at a point where legacy controls have little or no visibility.

Breaches seen in the wild

• Secrets in VS Code extensions 2025: Wiz found 550+ secrets in VS Code extensions, including publishing tokens able to push malicious updates to about 150,000 installs.
• Dropbox Sign breach 2024: A compromised back-end service account gave attackers Dropbox Sign customer data, including API keys, OAuth tokens and MFA information.

Read and download The State of NHI & AI Agent Breach Report 2026, covering 200+ breaches impacting Non-Human Identities including AI Agents.

NHI Mgmt Group analysis

Browser-native governance is now an identity issue, not a UX issue. The browser has become the place where authenticated state is reused, mutated, and exploited after the IdP has already handed off responsibility. That means identity programmes that end at login are structurally incomplete. Practitioners should treat browser telemetry and session behaviour as part of the identity control surface, not a separate endpoint problem.

Session memory is the new blast radius. Once tokens, cookies, and SaaS tabs coexist in the browser, a compromise no longer needs a password to matter. A single browser session can carry enough context to impersonate a user, cross application boundaries, and leak data through ordinary work activity. The governance implication is that risk is no longer bound to the account alone; it is bound to the active session state.

Invisible AI endpoints create a control gap across human, NHI, and autonomous behaviour. The same browser that supports human users is now also the conduit for GenAI prompts, browser assistants, and extension-driven automation. Those interactions collapse cleanly separated assumptions about who is acting, what is being shared, and where policy enforcement begins. The implication is that identity governance has to account for browser-mediated delegation chains, not just named users.

Browser extension trust debt is a supply-chain problem inside identity operations. Extensions inherit access to session material without being managed like software dependencies, which means the security team often has no inventory discipline, provenance logic, or update trust model. That gap is operational, not theoretical. Practitioners should govern extensions with the same seriousness they apply to third-party integrations and privileged software paths.

From our research library:

• Only 5.7% of organisations have full visibility into their service accounts, according to the Ultimate Guide to NHIs.
• Read next: Identity Visibility and Intelligence Platforms (IVIP) Guide

What this signals

Browser-native visibility now belongs in identity governance. Identity programmes that only observe the login event will keep missing where risk actually accumulates, which is inside the active browser session. The operating model has shifted, so the control plane has to shift with it.

Session drift is the governance problem practitioners need to watch. Users can move from sanctioned access into personal accounts, AI tools, and unmanaged extensions without a new authentication event, which makes post-login behaviour the decisive control point. That is why session-state monitoring matters more than a checkbox view of SSO coverage.

Invisible AI endpoints change the programme design question. The issue is no longer whether employees can use GenAI, but whether the browser path is governed when they do. Browser-native controls, extension oversight, and prompt-level DLP now sit in the same conversation as IAM and data loss prevention.

For practitioners

• Treat the browser as a primary control plane Move visibility and policy enforcement into browser activity so prompts, copy/paste, uploads, and account context are monitored where exfiltration actually occurs.
• Extend session governance beyond the IdP Track active sessions for token reuse, account crossover, and unmanaged browser access instead of assuming MFA and SSO end the risk at authentication.
• Govern browser extensions as supply-chain software Inventory extensions continuously, score publisher trust, review update cadence, and restrict sideloading where extensions can access cookies or SaaS tabs.
• Apply DLP to prompt and clipboard behavior Classify data in motion at the point of interaction and block risky copy/paste, drag-and-drop, and prompt-entry paths before content leaves the tab.
• Separate sanctioned AI use from unmanaged browser AI Provide approved AI and SaaS workflows with clear guardrails so users have a governed path that reduces pressure to paste sensitive material into personal tools.

Key takeaways

• The browser has become the place where identity, data movement, GenAI use, and extension risk converge outside traditional control boundaries.
• Most of the practical exposure comes after login, which means browser sessions and prompts are more important to govern than the authentication event alone.
• Teams that extend visibility into browser activity can reduce exfiltration risk without relying solely on IdP controls, file DLP, or endpoint tooling.

Key terms

• Browser DLP: Browser DLP is policy enforcement applied to web sessions and browser-based uploads. It matters because SaaS apps, webmail, and generative AI tools now act as primary data exit points, so organisations need controls that can inspect and stop transfers in the browser, not only in backend gateways.
• Session State: The information that persists across an interaction, including prompts, outputs, files, and intermediate context. In agentic systems, session state is a governance boundary because persistence, transfer, and reconstruction determine whether the conversation is auditable and whether data leaves the original trust domain.
• Extension provenance: Extension provenance is the trust history behind a browser add-on, including who published it, how it is updated, and whether it was sideloaded. In practice, it determines whether the extension should be treated as managed software or as an ungoverned supply-chain dependency.
• Identity Control Plane: An identity control plane is the governance layer that decides who or what can access systems and under what conditions. In practice, it coordinates authentication, authorization, privilege review, and lifecycle management across human and machine identities so access policy is enforced consistently across environments.

Deepen your knowledge

NHI governance, agentic AI identity, and machine identity lifecycle are core topics in our NHI Foundation Level course, the industry's only accredited NHI security programme. If you are building or maturing an IAM programme, it is worth exploring.