GitHub published six Chartbrew advisories between July and September 2026, and their CVE records went live on 6 November 2026. Chartbrew is an open-source tool that connects directly to a team’s databases and APIs to build charts. A flaw in its own login or access checks therefore reaches every data source it touches.
The six records cover a default signing secret, a SQL injection path and a cross-tenant project takeover. The remaining three cover a server-side request forgery, a gap in an AI query filter and a dashboard brute-force path. None of the six appear in the CISA Known Exploited Vulnerabilities catalogue, version 2026.11.06.
What the Chartbrew CVE Bundle Covers
CVE-2026-85295 is the most serious item. GitHub’s advisory says four of Chartbrew’s own login checks accept a session token signed with either of two keys. Real logins use only one key; the second, called CB_SECRET, ships in the project’s example setup file as a fixed placeholder value under a comment that calls it unused.
The advisory says a deployment that keeps that placeholder, or any other weak value, lets an outside caller forge a token for any account. That forged token can skip the password and two-factor steps. NVD and the advisory both score it CVSS 3.1 9.8 Critical.
CVE-2026-54349 is a SQL injection path in how Chartbrew fills variables into a saved query for public and embedded charts. The advisory says the substitution step does not fully account for a database engine’s own special-character handling. Because of that gap, a visitor who never logs in can change the query through a URL parameter. NVD scores it CVSS 3.1 9.8 Critical, the same as the forged-token flaw.
CVE-2026-85712 lets a registered user take over another team’s project. The advisory says a shared permission check reads the caller’s team ID from the request body instead of the project’s own team. Because every new signup owns a team by default, any user can supply their own ID and pass the check.
NVD scores it CVSS 3.1 8.8 High. This breaks the least privilege principle: a shared permission check must confirm both sides of a relationship, not just one.
CVE-2026-68935 is a server-side request forgery. The advisory says a user who can manage API connections can turn on an internal allowPrivateHost setting. That setting lets the Chartbrew server reach private network addresses it would otherwise refuse. An upgrade script also turned this setting on for every connection that already existed.
NVD scores it CVSS 3.1 9.1 Critical. Recognized cloud metadata addresses stay blocked even when the setting is on, per the advisory.
CVE-2026-69128 only affects instances that turn on Chartbrew’s AI query assistant. The advisory says the assistant’s PostgreSQL safety filter misses several file-reading functions and can be slipped past by quoting a blocked function name. That gap lets a team owner or admin read files from the connected database host. NVD lists only a CVSS 4.0 score of 4.6 Medium for this one; it records no CVSS 3.1 score at all.
CVE-2026-68933 covers password-protected public dashboards. The advisory says Chartbrew stored these passwords as plain text, accepted them through the page’s own URL, and checked them with no limit on repeated attempts. An outside attacker could use that gap to guess a dashboard password by sending enough requests. NVD scores it CVSS 3.1 7.5 High.
Why the Chartbrew CVE Bundle Landed All at Once
The six GitHub advisories were not all written on the same day. The SQL injection advisory first went public in July 2026, and the other five followed in September. The fixes themselves shipped between 14 June and 23 July 2026, across Chartbrew 5.2.1, 5.2.3 and 5.3.0. The CVE records only caught up to that work on 6 November 2026.
A team already running Chartbrew 5.3.0 or later from the summer was likely already protected months before this week’s CVE wave. A team still on an older release has six fixes to apply at once, not six separate incidents spread across the year. Exploiting a public-facing application this way needs no stolen credentials for five of the six flaws. That is why the version check below comes first.
The Order to Fix the Chartbrew CVE Bundle
Upgrading first closes the most paths at once, since Chartbrew’s own 5.2.3 change log says it rejects session tokens signed with the old CB_SECRET placeholder outright. Rotating secrets and checking for prior abuse still follow the upgrade, because a new version stops new forgeries without undoing anything an attacker already did with an old one.
- Find your version and exposureCheck whether your instance predates 5.2.1, 5.2.3 or 5.3.0, whether CB_SECRET still holds its shipped placeholder value, and whether the AI assistant or any API connections are turned on.
- Upgrade to the current releaseMove to Chartbrew 6.0.0 or later. GitHub’s advisories name a patched release for five of the six flaws (5.2.1, 5.2.3 or 5.3.0); Chartbrew’s own 5.2.3 change log closes the sixth. Every later release carries all six forward.
- Regenerate CB_SECRET and CB_ENCRYPTION_KEYSet new values for both keys after upgrading. The fix stops the server from accepting the old placeholder going forward. It does not cancel a token an attacker already holds.
- Review teams, projects and connectionsLook for a project that moved to a team you do not run, a dashboard that turned public on its own, and an API connection with allowPrivateHost turned on. The migration enabled that setting on every connection that already existed.
- Reset every public dashboard passwordChange the password on each protected dashboard once you upgrade. The old values were stored and sent as plain text, so treat them as already known.
- Re-check the AI assistant before you trust it againConfirm your release carries the Postgres and ClickHouse query guards from the 5.2.3 change log before you rely on the blocklist for a connected database.
What Brandefense Blackhole Shows About the Chartbrew CVE Bundle
Brandefense Blackhole has logged ten posts naming Chartbrew on dark-web and chat channels since 9 August 2026. Three of them, including one naming CVE-2026-85295 and two naming CVE-2026-54349, were posted on 6 November 2026, the same day the CVE records went live.
The six scores span a wide range: 9.8, 9.8, 9.1, 8.8 and 7.5, plus 4.6 on CVSS 4.0 alone for the AI-only flaw. That one has no CVSS 3.1 score in the record. Treat the bundle as six separate fixes to verify, not one severity rating to react to once.
Start with the version check above, since it tells you which of the six still apply. For the patch management step that follows, plan the secret rotation and password reset for the same maintenance window as the upgrade.