Tag
OpenAI describes security safeguards against URL-based data exfiltration attacks when AI agents retrieve web content, using an independent web index to verify that URLs are publicly known before automatic retrieval to prevent prompt injection attacks from leaking sensitive user data.