Home Content News Spring AI Flaw Adds New Risk To AI-Enabled Open Source Apps

Spring AI Flaw Adds New Risk To AI-Enabled Open Source Apps

0
1
Broadcom
Broadcom

Broadcom’s 91 Spring AI CVEs expose the downstream scale of open-source dependencies, with Sonatype estimating 209,569 affected software components and urging targeted remediation.

Broadcom released a batch of Spring security advisories on 20 August 2026, with Sonatype tracking 91 CVEs across Spring Framework and related open-source projects. The disclosure is estimated to affect 209,569 software components, highlighting how vulnerabilities can propagate through open-source software supply chains.

The affected projects include Spring Security, Spring Cloud Config, Spring AI, Spring Data REST, Spring Integration, Reactor Core, Reactor Netty, Spring AMQP, and Spring Batch. Vulnerable code can reach applications through transitive dependencies, embedded libraries, and upstream frameworks, meaning an upstream fix does not immediately secure downstream software.

The flaws include insecure deserialisation, untrusted code execution, information disclosure, server-side request forgery, path traversal, denial-of-service, and authorisation weaknesses. CVE-2026-59285, affecting Spring for GraphQL, carries a 9.2 Critical rating from Sonatype and could potentially enable remote code execution under specific Jackson and GraphQL configurations.

CVE-2026-59318 affects Spring AI tool calling, where prompt injection could trigger an unintended tool and potentially enable privilege escalation if permissions are not enforced at the system layer.

Sonatype said newly affected component versions increased at 46 times the pre-AI rate, while critical and high-severity vulnerabilities per enterprise application rose 4.31 times over its four-year analysis.

Organisations should prioritise internet-facing services, GraphQL deployments, Spring AI implementations, and applications handling untrusted data. Sonatype recommends identifying direct and transitive dependencies, scanning SBOMs, testing fixed versions, and prioritising vulnerabilities by reachable attack paths rather than treating all 91 CVEs equally.

Loading form…

LEAVE A REPLY

Please enter your comment!
Please enter your name here