Developer

Metabase Discloses Critical Zero-Day SQL Injection Flaw With Perfect CVSS Score

Metabase disclosed a critical zero-day SQL injection vulnerability (CVE-2026-72898) with a perfect CVSS score of 10, affecting versions 1.58 and up. The flaw allows attackers raw SQL access via the password reset endpoint, impacting companies like Kilo Code, Tally, Framework, n8n, and ChecklyHQ. Metabase has patched the issue and urges self-hosted users to upgrade immediately.

Neura News

Neura News

Neura Market Editorial

August 12, 20265 min read
Metabase Discloses Critical Zero-Day SQL Injection Flaw With Perfect CVSS Score

Metabase disclosed a critical zero-day SQL injection vulnerability on August 6, designated CVE-2026-72898, that gives attackers raw SQL access to underlying databases and has already hit multiple companies. The flaw carries a severity score of 10, the highest possible rating on the CVSS scale, and affects Metabase versions 1.58 and up. Several organizations, including Kilo Code, Tally, Framework, n8n, and ChecklyHQ, reported being impacted.

A Perfect Score and an Old School Attack

The vulnerability allows attackers to inject SQL through the /api/session/reset_password endpoint, a path meant for password recovery. David Shipley, CEO of Beauceron Security, said the attack method is "old school and painful" and noted that working proof of concept exploit code exists. "You don't see a perfect 10/10 on CVSS often, but when you do, be worried," Shipley said.

The attack pattern is specific and detectable. It involves a POST /api/session/reset_password request that returns a 400 status code, followed by a GET /api/user/current request that returns a 200 status code. Metabase warned customers about this pattern in its disclosure. "If you find that pattern in your application logs or in your Metabase server ingress logs, it is likely that your instance has been compromised," the company said.

Scott Miserendino, CTO at DataBee, called the flaw "a very serious vulnerability." He said it allows attackers to have "unmitigated, raw SQL access" to the Metabase database. From there, attackers can steal or alter account credentials for connected databases, create new admin accounts, change app configurations, escalate privileges, or "degrade, alter or destroy" information.

Broad Reach Across Databases and Deployments

Metabase is an open-source business intelligence tool used for analytics, querying, and visualizing data. It can connect to a wide range of databases, including Databricks, MongoDB, Oracle, Snowflake, Amazon, BigQuery, and others. That broad connectivity means a single compromised instance can expose many backend systems.

The exposure is significant. Shodan, a search engine for internet-connected devices, tracked roughly 2,500 Metabase instances. Wiz, a security company, reported that about 13% of cloud environments have deployed self-hosted Metabase instances. Of those, about 25% are fully accessible on the internet, meaning a quarter of self-hosted deployments are directly reachable by anyone.

Metabase Cloud customers were already upgraded and patched, according to the company. Self-hosted customers may still be vulnerable unless they apply the patch. The company has blocked exploited endpoints, patched the vulnerability, terminated sessions, and revoked credentials.

Miserendino noted that the vulnerability can also affect platforms using OEM versions of Metabase, which are embedded in other products. Affected users may not even know they are affected, he said.

Impacted Companies and What Was Exposed

The list of impacted companies includes Kilo Code, which was recently acquired by Anaconda, Tally, a Y Combinator backed startup building autonomous accounting agents, Framework, a personal computer manufacturer, n8n, a workflow automation platform, and ChecklyHQ, an AI testing and monitoring platform provider.

Threat actors accessed records containing usernames, email addresses, cloud passwords, cryptographic hashes of OpenTelemetry (OTel) API keys, Slack access tokens, and other sensitive information. OTel API keys are used for trace collection in observability systems, and Slack tokens are used for authentication in Slack integrations.

The #1 Newsletter in AI

Stay ahead of the AI curve

The most important updates, news, and content — delivered weekly.

No spam. Unsubscribe anytime.

The impacted companies are contacting customers and taking mitigation actions. Those actions include rotating credentials and API keys, resetting user passwords, invalidating Slackbot authentication tokens, removing compromised admin accounts, and reviewing audit logs.

So far, the impacted companies appear to be smaller organizations and startups, according to the disclosure.

Checkly's Response and a Broader Lesson

Checkly issued a notice about the incident and said it is rethinking internal processes around analytics tools. The company is improving sanitation, performing audits, and paging on-call engineers for future vendor security notices.

"The vulnerability was in a vendor's product, but protecting your data is our job, and this incident put some of it at risk," Checkly said in its notice. The company acknowledged that immediate fixes are not enough. "Rotating credentials fixes the immediate problem. It does not fix the reason this hurt: Our analytics environment held more sensitive data and had broader access than it needed."

Anaconda, which acquired Kilo Code, urged customers to remain diligent. The company said to "be on alert for phishing/social engineering, and maintain credential hygiene (including auditing, reviewing, and rotating credentials regularly) and spam monitoring."

Patching and Mitigation Guidance

Metabase recommends upgrading to the patched version. For example, if an organization is running version 0.58.6, it should move to 0.58.24 or later. A temporary workaround is to block the /api/session/reset_password endpoint.

If the endpoint was publicly accessible, enterprises should take more extensive steps. Metabase advises revoking all active user sessions, reviewing and deleting unrecognized API keys, auditing data warehouse logs and admin accounts, rotating credentials for all connected databases, and reviewing Metabase activity and query histories.

Miserendino offered additional advice for enterprises. He recommends wrapping third-party REST interfaces with their own SQL injection detection, using a web access firewall or reverse proxy. He also said enterprises should monitor their security and database logs for the creation of new or recently elevated administrator accounts, or other unusual activity such as large volumes of data drops.

The disclosure comes amid a busy period for security teams. A related article on a Ruby on Rails critical bug was published on August 4, 2026, just two days before the Metabase disclosure. The Metabase article originally appeared on CSOonline.

Related on Neura Market

More from Neura News

Industry

Google Cuts Pixel 11 Pro AI Trial to Six Months, Adds Three Costly Catches

Google has reduced the free Google AI Pro trial bundled with the Pixel 11 Pro from 12 months to six months, cutting the perk's value by $119.94. The change applies across the Pixel 11 Pro lineup and introduces three costly catches, including losing the trial if upgrading to AI Ultra, auto-renewal before the next flagship launch, and termination of existing promos when redeeming new ones. The Pixel 10 Pro still offers the full 12-month trial, making it a viable alternative for shoppers.

Aug 16·4 min read
Research

LittleLearner Models Trained Only on K-5 Curriculum Show Skills Are Elicited, Not Acquired

Researchers released LittleLearner, a family of language models trained from scratch on a strictly filtered K-5 elementary school curriculum, to answer whether capabilities beyond training data can be elicited or acquired through scaling, post-training, and in-context learning. The answer is largely no: scaling, post-training, and in-context learning amplify what the curriculum taught, but none meaningfully improve out-of-scope performance. The pretraining filter sets the effective capability ceiling, providing a controlled sandbox for studying knowledge acquisition and RL.

Aug 16·5 min read
Industry

The Hidden Gold Rush: Scammers Exploit Demand for Claude Watermark Removal Apps

Anthropic's August 2026 watermarking of Claude text has sparked a surge in demand for removal apps, attracting scammers who peddle fraudulent tools. AI scientist Lance Eliot warns these apps often contain malware or fail to work, as statistical watermarks are nearly impossible to remove without heavy editing. With billions of users at risk, the problem is expected to worsen as more AI makers adopt watermarking.

Aug 16·12 min read