A large language model called WormGPT and branded by its developer as a ChatGPT alternative with "no ethical boundaries or limitations" may already placing malicious outputs in your inbox.

The LLM, built on the open source GPT-J model, was designed to generate sleek, polished phishing emails – allowing threat actors to launch convincing BEC attacks, say security researchers.

Get the full story: Subscribe for free

Get the story, a weekly newsletter (you can turn that off if you want) and help us fight bots and feral algorithms. Subscribe today.

Subscribe now

Already a member? Sign in