Prisma Browser EUVD-2026-30062

Requires local interactive or shell access to a macOS endpoint running a vulnerable Prisma Browser build (<146.16.6.165) as a non-admin authenticated user; no admin rights, no user interaction from a victim, and no network reachability are needed. Limiting factors: the attacker must already have a valid local account on the device (so this is an insider/post-compromise privilege boundary issue, not a remote vector), only macOS installations are affected, and the AT:P (attack requirements: present) flag plus SSVC 'not automatable' indicate the exposed bridge must be in its normal running state - meaning the browser has to be installed and active - rather than triggerable on arbitrary unrelated hosts.

The CVSS 4.0 vector (AV:L/AC:L/AT:P/PR:L/UI:N) confirms this is local, low-complexity, requires a present attacker condition, and needs low (non-admin) authentication - consistent with the description's 'locally authenticated non-admin user' wording, with no signal conflict. Confidentiality, integrity, and availability impact to the vulnerable system are all High, and SSVC rates technical impact as 'total' which matches a full bypass of the browser's security model. Counter-signals lower the urgency: EPSS is 0.01% (0th percentile), SSVC exploitation status is 'none', not automatable, and the CVE is not in CISA KEV, so this is a credible local-escalation issue against a security product rather than a mass-exploitation concern. Priority should be elevated for organizations that deploy Prisma Browser specifically as a DLP/SaaS-access control layer on shared or multi-user macOS endpoints, since the entire premise of that control is what the bug bypasses.

A contractor or standard (non-admin) user logged into a corporate macOS device opens a local helper script that connects to Prisma Browser's automation bridge and issues commands to navigate the browser, exfiltrate session data, or disable in-browser DLP controls - actions the policy layer would normally block. Because the bridge accepts the commands without the expected trust check, the user effectively impersonates the browser's privileged automation client and bypasses the security controls the enterprise relies on. No public exploit is identified at time of analysis, but the conditions are simple enough that any insider with local shell access could develop one.