This research provides the first comprehensive safety risk assessment of AI-Powered Search Engines (AIPSEs) that combine LLMs with external databases.

• AIPSEs are vulnerable to quoting malicious content and citing harmful websites
• Seven production AI search engines were evaluated for security weaknesses
• Researchers identified specific attack vectors including phishing attempts and harmful content dissemination
• Defensive measures and mitigation strategies are proposed to protect users

This work is crucial for security professionals and AI developers as it highlights how traditional web security risks are evolving in next-generation search technologies, requiring new protective approaches.

The Rising Threat to Emerging AI-Powered Search Engines