Skip to content
Weaponized · 1 source ·
Picture from the securelist.com article
Image: securelist.com

Phishing pages written by LLMs keep confessing 'As an AI language model' in the copy

Ourtake

The scammers automated the scam pages, then forgot to automate the proofreading.

What happened

Securelist researchers examined fake crypto exchange and wallet login pages used in phishing campaigns. The pages were generated with LLMs and kept telltale artifacts such as apologies, refusals and knowledge-cutoff notes. The pages impersonated brands including Crypto.com, KuCoin, MetaMask and Coinbase to collect credentials.

Source

Report a mistake