Improper LLM Output Handling: SQL Injection, XSS, and SSRF via AI-Generated Responses
LLM output is attacker-influenced data. Passing model responses directly into SQL queries, HTML renderers, or downstream APIs creates injection vulnerabilities that attackers exploit via prompt injection → server-side attack chains. Concrete defenses for every consumer type.