CVE-2026-59369

Authorization Bypass
Affects
Spring AI
in
Spring
No items found.
Versions
>=1.0.0 <=1.0.10, >=1.1.0 <=1.1.9, >=2.0.0 <=2.0.2

Patch Available.

Exclamation circle icon
Patch Available

This Vulnerability has been fixed in the Never-Ending Support (NES) version offered by HeroDevs.

Overview

Spring AI is a Spring project that provides a Spring-friendly API and abstractions for developing AI applications, applying Spring ecosystem design principles such as portability and modular design to the AI domain, including portable integrations with vector stores such as Typesense.

An authorization bypass vulnerability (CVE-2026-59369) has been identified in the Spring AI Typesense Vector Store, which allows attackers to inject additional clauses into metadata filter expressions through a crafted filter value. An injected clause can override application-enforced restrictions, such as tenant or permission scoping filters, and disclose documents the attacker should not be able to see.

Per OWASP: An injection flaw is a vulnerability which allows an attacker to relay malicious code through an application to another system.

This issue affects multiple versions of Spring AI.

Details

Module Info

Vulnerability Info

This High-severity vulnerability is found in the org.springframework.ai:spring-ai-typesense-store package in multiple versions of Spring AI.

TypesenseVectorStore converts portable Filter.Expression objects, such as those passed to SearchRequest.filterExpression(...) or used by filter-based deletes, into a Typesense filter_by string with TypesenseFilterExpressionConverter. Each metadata key becomes metadata.<key>:, logical operators become && and ||, and each value is written by doSingleValue:

@Override
protected void doSingleValue(Object value, StringBuilder context) {
    emitJsonValue(value, context);
}

emitJsonValue serializes the value as a JSON literal, so a string becomes a double-quoted JSON string with JSON escapes. Typesense's filter_by grammar is not JSON: only backticks delimit a string literal, and a double quote is matched as part of the value rather than as a delimiter. Characters inside the value such as &&, ||, ( and ) are therefore parsed as filter syntax. When an application builds a filter that combines its own restriction (for example tenant == 'acme') with a value taken from untrusted input, an attacker can supply a value that closes the intended clause and appends an additional || condition of their choosing. The resulting filter_by no longer enforces the application's restriction, so searches return documents from other tenants or permission scopes. Delete operations that take a filter expression go through the same conversion.

Mitigation

Only recent versions of Spring AI receive community support. Older lines are End-of-Life and will not receive public updates to address this issue.

Users of the affected components should apply one of the following mitigations:

  • Upgrade to a currently supported version of Spring AI.
  • Leverage a commercial support partner like HeroDevs for post-EOL security support.

Credits

Vulnerability Details
Severity
Level
CVSS Assessment
Low
>=0 <4
Medium
>=4 <6
High
>=6 <8
Critical
>=8 <10
High
ID
CVE-2026-59369
PROJECT Affected
Spring AI
Versions Affected
>=1.0.0 <=1.0.10, >=1.1.0 <=1.1.9, >=2.0.0 <=2.0.2
NES Versions Affected
Published date
November 19, 2026
≈ Fix date
Category
Authorization Bypass
Vex Document
Download VEXHow do I use it?
Sign up for the latest vulnerability alerts fixed in
NES for Spring
Rss feed icon
Subscribe via RSS
or

By submitting the form I acknowledge receipt of our Privacy Policy.

Thanks for signing up for our Newsletter! We look forward to connecting with you.
Oops! Something went wrong while submitting the form.