Dark GPT refers to malicious or jailbroken AI language models, such as WormGPT and FraudGPT, that have had their safety guardrails removed so…
The term covers two related things.
Most underground models are not trained from scratch — that would cost millions of dollars and require infrastructure few criminals have.
The most consistent real-world use is social engineering.
A lot of the coverage frames dark GPT as a mysterious, all-powerful AI that companies are hiding.
Read the complete breakdown, fixes and what happens next.