An AI red-teaming company found that xAI’s Grok 4 is “not suitable for enterprises” without substantial security prompting. 

A pair of AI tools advertised on hacking forums were developed using commercial AI models from xAI and Mistral, according to Cato Networks.