In Spring AI, a SpEL injection vulnerability exists in SimpleVectorStore when a user-supplied value is used as a filter expression key. A malicious actor could exploit this to execute arbitrary code. Only applications that use SimpleVectorStore and pass user-supplied input as a filter expression key are affected. This issue affects Spring AI: from 1.0.0 before 1.0.5, from 1.1.0 before 1.1.4.

A critical SQL injection vulnerability in Spring AI's MariaDBFilterExpressionConverter allows attackers to bypass metadata-based access controls and execute arbitrary SQL commands.

The vulnerability exists due to missing input sanitization.

SQL injection vulnerability in Spring AI's CosmosDBVectorStore allows attackers to execute arbitrary SQL queries via crafted document IDs.

Affected versions: Spring AI: 1.0.0 - 1.0.5 (fixed in 1.0.6), 1.1.0 - 1.1.4 (fixed in 1.1.5)

A JSONPath injection vulnerability in Spring AI's AbstractFilterExpressionConverter allows authenticated users to bypass metadata-based access controls through crafted filter expressions. User-controlled input passed to FilterExpressionBuilder is concatenated into JSONPath queries without proper escaping, enabling attackers to inject arbitrary JSONPath logic and access unauthorized documents.

This vulnerability affects applications using vector stores that extend AbstractFilterExpressionConverter for multi-tenant isolation, role-based access control, or document filtering based on metadata.

The vulnerability occurs when user-supplied values in filter expressions are not escaped before being inserted into JSONPath queries. Special characters like ", ||, and && are passed through unescaped, allowing injection of arbitrary JSONPath logic that can alter the intended query semantics.

Spring AI's spring-ai-bedrock-converse contains a Server-Side Request Forgery (SSRF) vulnerability in BedrockProxyChatModel when processing multimodal messages that include user-supplied media URLs. Insufficient validation of those URLs allows an attacker to induce the server to issue HTTP requests to unintended internal or external destinations. This issue affects Spring AI: from 1.0.0 before 1.0.5, from 1.1.0 before 1.1.4.

In Spring AI, various FilterExpressionConverter implementations accept a filter expression object and translate them to specific vector store query languages. In several cases, keys and values are not properly escaped, leading to the ability to alter the query.

Affected versions: Spring AI: 1.0.0 - 1.0.5 (fixed in 1.0.6), 1.1.0 - 1.1.4 (fixed in 1.1.5)

Spring AI's MilvusVectorStore#doDelete(List) implementation is vulnerable to filter-expression injection via unsanitized document IDs. Spring AI 1.0.x: affected from 1.0.0 through latest 1.0.x; upgrade to 1.0.7 or greater. Spring AI 1.1.x: affected from 1.1.0 through latest 1.1.x; upgrade to 1.1.6 or greater.

In Spring AI Vector Stores, special characters could be used to force the execution of arbitrary queries in Elasticsearch, OpenSearch, and GemFire VectorDB. Affected components: spring-ai-elasticsearch-store, spring-ai-opensearch-store, spring-ai-gemfire-store.

Affected versions: Spring AI 1.0.0 through 1.0.x (fix 1.0.9). Spring AI 1.1.0 through 1.1.x (fix 1.1.8).

A malicious user could craft input that is stored in conversation memory and later interpreted by the model in an unintended way. Applications using the affected advisor with user-controlled input may be susceptible to manipulation of model behavior across conversation turns.

Spring AI's spring-ai-neo4j-store contains a Cypher injection vulnerability in Neo4jVectorFilterExpressionConverter. When a user-controlled string is passed as a filter expression key in Neo4jVectorFilterExpressionConverter of spring-ai-neo4j-store, doKey() embeds the key into a backtick-delimited Cypher property accessor (node.metadata.) after stripping only double quotes, without escaping embedded backticks.This issue affects Spring AI: from 1.0.0 before 1.0.5, from 1.1.0 before 1.1.4.

In RedisFilterExpressionConverter of spring-ai-redis-store, when a user-controlled string is passed as a filter value for a TAG field, stringValue() inserts the value directly into the @field:{VALUE} RediSearch TAG block without escaping characters.This issue affects Spring AI: from 1.0.0 before 1.0.5, from 1.1.0 before 1.1.4.

Spring AI's chat memory component contained a problematic default that, when not explicitly overridden, could result in unintended data exposure between users.

Analyzing a PDF with a deeply nested or cyclic table of contents can cause a StackOverflowError in the ingestion thread. Spring AI 2.0.0 Spring AI 1.1.0 - 1.1.8 Spring AI 1.0.0 - 1.0.9

A local attacker on a multi-user host can pre-create the deterministic cache path and plant a malicious ONNX model file. Spring AI 2.0.0 Spring AI 1.1.0 - 1.1.8 Spring AI 1.0.0 - 1.0.9

In Spring AI, a malicious PDF file can be crafted that triggers the allocation of unreasonable amounts of memory when handled by ForkPDFLayoutTextStripper.

Affected versions: Spring AI: 1.0.0 - 1.0.5 (fixed in 1.0.6), 1.1.0 - 1.1.4 (fixed in 1.1.5)

Spring AI's support for Anthropic's Skills API used LLM-influenced filenames unsanitized in Path.resolve before writing files to disk. This could allow a malicious user to write files outside the intended target directory, including restricted directories.

Affected versions: Spring AI: 1.1.0 through 1.1.x

ResourceCacheService.getCacheName() builds the on-disk filename by appending the URI fragment verbatim, without stripping path separators or .. sequences, and passes the result to new File(resourceParentFolder, newFileName) before writing the downloaded bytes there. Spring AI 2.0.0 Spring AI 1.1.0 - 1.1.8 Spring AI 1.0.9 and earlier

In Spring AI, having access to a shared environment can expose the ONNX model used by the application.

Affected versions: Spring AI: 1.0.0 - 1.0.5 (fixed in 1.0.6), 1.1.0 - 1.1.4 (fixed in 1.1.5)

In Spring AI, an attacker can bypass conversation isolation and exfiltrate sensitive memory from other users’ chat histories, including secrets and credentials, by injecting filter logic through conversationId. Only applications that use VectorStoreChatMemoryAdvisor and pass user-supplied input as a conversationId are affected.

RedisChatMemoryRepository.findByMetadata() builds RediSearch tag and text queries from caller-supplied metadata values without applying RediSearchUtil.escape(), unlike get(), clear(), and findByTimeRange() in the same class which do escape their inputs. An application that passes user-controlled values to findByMetadata() on a tag-typed metadata field allows an attacker to inject RediSearch syntax (e.g. x} | ) that breaks out of the tag clause and matches all indexed chat messages across every conversation in the index. Spring AI 2.0.0