CVE-2026-32628
AnythingLLM is an application that turns pieces of content into context that any LLM can use as references during chatting.
Executive Summary
CVE-2026-32628 is a high severity vulnerability affecting appsec, ai-code. It is classified as SQL Injection. Ensure your systems and dependencies are patched immediately to mitigate exposure risks.
Precogs AI Insight
"Architecturally, this flaw occurs due to within AnythingLLM, allowing the absence of comprehensive security boundaries. This flaw provides a direct pathway for attackers to inject malicious logic that alters the execution flow of the application engine. By intercepting insecure data flows from user input directly to rendering sinks, Precogs is designed to alert security teams to imminent boundary violations."
What is this vulnerability?
CVE-2026-32628 is categorized as a critical SQL Injection flaw. Based on our vulnerability intelligence, this issue occurs when the application fails to securely handle untrusted data boundaries.
AnythingLLM is an application that turns pieces of content into context that any LLM can use as references during chatting. In 1.11.1 and earlier, a SQL in...
This architectural defect enables adversaries to bypass intended security controls, directly manipulating the application's execution state or data layer. Immediate strategic intervention is required.
Risk Assessment
| Metric | Value |
|---|---|
| CVSS Base Score | 8.8 (HIGH) |
| Vector String | CVSS:3.1/AV:N/AC:L/PR:L/UI:N/S:U/C:H/I:H/A:H |
| Published | March 16, 2026 |
| Last Modified | March 16, 2026 |
| Related CWEs | CWE-89 |
Impact on Systems
✅ Data Exfiltration: Full compromise of the database schema, allowing extraction of all tables, user records, and PII.
✅ Authentication Bypass: Attackers can manipulate boolean logic in authentication queries to log in as administrators.
✅ Remote Code Execution: In severe configurations (e.g., xp_cmdshell in MSSQL), attackers can execute shell commands on the database underlying OS.
How to fix this issue?
Implement the following strategic mitigations immediately to eliminate the attack surface.
1. Prepared Statements Migrate entirely to parameterized queries (Prepared Statements) or an Object-Relational Mapper (ORM) to decouple code from data.
2. Input Validation Implement rigorous allow-list input validation for all sorting, filtering, and query parameters.
3. Principle of Least Privilege Ensure the database service account has the minimum necessary privileges, restricting DROP, TRUNCATE, and system execution commands.
Vulnerability Signature
// Example of a vulnerable Node.js/Express snippet
const category = req.query.category;
// DANGEROUS: Direct string concatenation of user input
const query = `SELECT * FROM products WHERE category = '$\{category\}'`;
db.query(query, (err, result) =\> \{
if (err) throw err;
console.log(result);
\});
// SECURED: Using parameterized queries avoids SQL injection
const category = req.query.category; // Ensure scope appropriately
// Safe: The database driver treats '?' strictly as data, not executable code
const query = 'SELECT * FROM products WHERE category = ?';
db.query(query, [category], (err, result) =\> \{
if (err) throw err;
console.log(result);
\});
References and Sources
- NVD — CVE-2026-32628
- MITRE — CVE-2026-32628
- CWE-89 — MITRE CWE
- CWE-89 Details
- Application Security Vulnerabilities
- AI Code Security Vulnerabilities
Vulnerability Code Signature
Attack Data Flow
| Stage | Detail |
|---|---|
| Source | User-controlled HTTP request parameter |
| Vector | String concatenation into SQL query string |
| Sink | Database engine executes the malformed query |
| Impact | Full database compromise, unauthorized data modification or exfiltration |
Vulnerable Code Pattern
# ❌ VULNERABLE: Direct string concatenation
def get_user(user_id):
query = "SELECT * FROM users WHERE id = '" + user_id + "'"
cursor.execute(query) # Taint sink: unparameterized query
return cursor.fetchone()
Secure Code Pattern
# ✅ SECURE: Parameterized query
def get_user(user_id):
query = "SELECT * FROM users WHERE id = %s"
cursor.execute(query, (user_id,)) # Sanitized binding
return cursor.fetchone()
How Precogs Detects This
Precogs AI Analysis Engine traces data flow from HTTP request parameters through string concatenation directly into database execution sinks, identifying critical SQL injection vectors via Code Property Graph traversal.\n