Keeping your data safe when an AI agent clicks a link

OpenAI Blog · 5 months ago

OpenAI has implemented safeguards to protect user data when AI agents open links by preventing URL-based data exfiltration and prompt injection attacks. The company uses techniques including sandboxing, input validation, and output filtering, though OpenAI does not publicly disclose specific technical thresholds or benchmark metrics for these protections. These measures reduce the risk that malicious URLs could extract sensitive information from user sessions or manipulate agent behavior through crafted link contents.