The Chinese company Moonshot, maker of the artificial intelligence models Kimi, has launched an internal investigation after investigators managed to force two of the company's most used models to provide information about the creation of biological weapons and assassination.

The company Mindguard, which tests the security of artificial intelligence systems, told the BBC in July that it discovered that two "chemical" models, the K2.6 and the K3 Swarm, managed to bypass the Yemeni restrictions imposed by their creators.

This problem was discovered during a process known as "jailbreaking" (removing restrictions); A method where researchers use sets of complex commands to try to determine whether or not AI tools can ignore the restrictions and obstacles Yemen has set for them.

The Chinese "Monshot" company announced in an interview with the BBC that it welcomes the feedback and participation of independent parties and considers it "one of the main pillars of building a better and more advanced artificial intelligence."