NextFin News - Google has officially released patches for two critical vulnerabilities in its Looker business intelligence and data analytics platform that could have enabled remote code execution (RCE) and the unauthorized exfiltration of sensitive internal metadata. According to SC Media, the flaws were discovered by security firm Tenable and disclosed on February 4, 2026, following a coordinated response with Google. The vulnerabilities, collectively referred to as "LookOut," posed a significant threat to the more than 60,000 organizations worldwide that rely on Looker for data modeling and enterprise reporting.
The first vulnerability involved a sophisticated RCE chain within the Looker Modeling Language (LookML) project environment. By exploiting a path traversal flaw in how Looker handles remote Git dependencies, researchers demonstrated that an attacker with developer permissions could bypass security controls to execute malicious code via Git hooks. This process required winning a race condition against Looker’s automated configuration overwrites, but once successful, it granted the attacker the ability to run commands on the host system. In cloud-hosted environments, this flaw carried the added risk of cross-tenant traversal, potentially allowing access to shared secrets folders belonging to other customers.
The second flaw, tracked as CVE-2025-12743, targeted Looker’s internal MySQL database, known as "looker__ilooker." This database serves as the platform's "brain," storing user credentials, permissions, and configuration secrets. According to Tenable, researchers bypassed UI restrictions by intercepting HTTP requests to force a connection to this internal database. By triggering intentional SQL query errors through Looker’s "data tests" feature, they were able to leak pieces of sensitive data through error messages, eventually allowing for the reconstruction of the entire database. Google has since updated its managed services, but organizations running self-hosted or on-premises versions must manually update to versions 25.12.30+, 25.10.54+, or higher to mitigate these risks.
From an analytical perspective, the "LookOut" vulnerabilities expose a fundamental tension in modern SaaS architecture: the trade-off between providing users with powerful, flexible development tools and maintaining strict security boundaries. Looker’s strength lies in LookML, which allows developers to treat data modeling like software engineering. However, by integrating deeply with Git and allowing remote dependencies, Google inadvertently expanded the attack surface. The use of Git hooks—a standard feature for automation—became a vector for code execution because the platform’s isolation mechanisms did not sufficiently account for path traversal in project naming conventions.
The cross-tenant risk identified in this case is particularly alarming for the broader cloud industry. As U.S. President Trump’s administration continues to emphasize the migration of federal data to commercial cloud providers, the integrity of multi-tenant isolation remains a top-tier national security concern. If an attacker can move from one tenant’s environment into a shared secrets directory, the foundational promise of cloud security—that your data is logically separated from your neighbor’s—is compromised. This incident serves as a reminder that even top-tier providers like Google are not immune to logic flaws that can bridge these gaps.
Furthermore, the data exfiltration method used in CVE-2025-12743 highlights a recurring weakness in complex web applications: the misuse of error reporting. By leveraging the "data tests" feature to leak internal database contents, the researchers utilized a functional business tool as a diagnostic weapon. This "error-based injection" is a classic technique, yet it remains effective in modern platforms because developers often prioritize detailed debugging information over strict data masking in specialized administrative modules. For financial institutions and healthcare providers using Looker, the potential leak of the "looker__ilooker" database could have resulted in the compromise of administrative credentials, leading to broader network penetration.
Looking ahead, the industry is likely to see a shift toward "Zero Trust" development environments where even internal developers are restricted from interacting with the underlying file system or internal management databases. We expect Google and its competitors to implement more aggressive sandboxing for data modeling languages and to move away from relying on local Git configurations that can be manipulated via path traversal. Additionally, the lag in patching for self-hosted customers remains a systemic risk. While Google Cloud customers were protected automatically, the thousands of firms running on-premises versions remain vulnerable until they navigate their internal change-management hurdles. This "patching gap" will continue to be a primary target for state-sponsored actors and cybercriminals who monitor disclosure reports to target slow-moving enterprise targets.
Explore more exclusive insights at nextfin.ai.
