Phantom Squatting: AI-Driven Supply Chain Threat Emerges

• •Phantom squatting involves registering AI-generated nonexistent domains for malicious use.
• •Unit 42's research identified 250,000 unique phantom domains from 685,339 queries.
• •Attackers can exploit the trust in LLM outputs, making detection challenging.

Cybercriminals are exploiting a new attack vector called 'phantom squatting,' where large language models (LLMs) generate plausible but nonexistent domains for legitimate brands. These hallucinated domains are then preemptively registered by attackers to host phishing kits and malware. Research from Palo Alto Networks' Unit 42 revealed that 685,339 queries against 913 global brands produced approximately 250,000 unique phantom domains, with over 13,220 confirmed malicious URLs. The attack lifecycle involves four phases: Discover, Act, Lure, and Bypass. Notably, the Montana Empire case demonstrated how a hallucinated domain was flagged as high-risk before being registered by an attacker. This attack vector poses a significant threat as it exploits the trust placed in LLM outputs, which lack historical telemetry, making them difficult to detect. The current status indicates ongoing exploitation of this vector, with attackers rapidly registering domains after LLM outputs.