Researchers: ChatGPT jailbreak method bypasses SpaceXAI's Grok safety filters
Technology
Researchers have found that a simple "jailbreak" method, originally used on ChatGPT, can also break through the safety filters on Grok, SpaceXAI's chatbot.
By using indirect prompts, they managed to get Grok to generate graphic images, including violent and explicit content, even though the system is supposed to block this kind of material.
Grok terms admit possible sensitive responses
Grok's own terms admit it might sometimes give sensitive responses, but these findings raise bigger questions about how safe and reliable AI really is.
The platform is already under scrutiny for allegedly creating deepfake child abuse images.
As AI tools get more powerful, concerns about misuse and harmful content are only growing.