Browser Agent

Expanded Definition

A browser agent is more than an automation script. In NHI security, it is an execution identity that can open sessions, follow redirects, submit forms, and act inside authenticated web workflows with the permissions and state of the account it is operating under. That makes it materially different from a headless scraper or a simple RPA job because it can inherit cookies, MFA-approved sessions, CSRF tokens, and other trust artifacts that were intended for human use. For governance purposes, browser agents sit at the intersection of application automation, session management, and Non-Human Identity controls. The term is still evolving across vendors, so definitions vary on whether the browser agent is the model, the automation runtime, or the delegated user context. NHI Management Group treats the browser agent as the full operational entity that executes browser-based actions under delegated authority, regardless of implementation stack. For broader context, the NIST AI Risk Management Framework is useful for framing autonomy, oversight, and misuse risk, while the OWASP Agentic AI Top 10 highlights the security impact of tool access and delegated execution. The most common misapplication is treating a browser agent as a harmless UI macro, which occurs when teams ignore its access to live sessions and production data.

Examples and Use Cases

Implementing browser agents rigorously often introduces session-control overhead, requiring organisations to balance task automation against tighter identity boundaries and monitoring.

• A support browser agent logs into a SaaS admin console to resolve routine account issues, but only through tightly scoped delegated access and time-limited session handling.
• A procurement agent navigates vendor portals to retrieve invoices, where the organisation must decide whether the efficiency gain justifies the risk of exposing payment workflows to an autonomous runtime.
• A QA agent submits forms and validates workflow behavior across authenticated pages, which makes traceability and action logging essential for post-incident review.
• An operations agent updates records in a browser-based internal system after receiving approval, illustrating why browser automation should be governed as an NHI rather than as a generic script.
• As discussed in OWASP NHI Top 10, browser-facing agents can expand blast radius when credentials, prompts, and session state converge in one execution path.

Why It Matters in NHI Security

Browser agents matter because they can turn a single stolen session or overly broad privilege grant into direct web application compromise. Once a browser agent is allowed to act like a person, it may bypass assumptions built into MFA, session timeout logic, and user behavior monitoring. That is why NHI governance must include identity scoping, step-up approval for sensitive actions, and explicit revocation when delegated access is no longer required. The risk is not theoretical: NHI Mgmt Group reports that only 5.7% of organisations have full visibility into their service accounts, which is a strong indicator that many browser agents will also operate without adequate inventory or oversight. Combined with the fact that 80% of identity breaches involved compromised non-human identities such as service accounts and API keys, browser agents become a high-value target whenever they inherit interactive trust. Related threat modeling guidance from CSA MAESTRO agentic AI threat modeling framework and MITRE ATLAS adversarial AI threat matrix helps teams map abuse paths. Organisations typically encounter the real danger only after a browser agent has been used in a phishing-assisted session theft or unauthorized transaction, at which point browser-agent controls become operationally unavoidable to address.

Related resources from NHI Mgmt Group

• AI Agent Authentication
• What breaks when a local AI agent service accepts browser connections from any website?
• Why do browser-native agent workflows increase identity risk?
• Who is accountable for actions taken by a browser agent inside an authenticated session?