
Artificial intelligence researchers have developed a novel technique to make ChatGPT and other popular chatbots safer.
The method, referred to as “neuron freezing”, prevents users from bypassing the built-in safety filters of the large language models (LLMs) underpinning these AI tools.