On September 17, Interface News learned that Ant AI Security Lab opened the SingProbe large model security fence and opened the code, models, and evaluation benchmarks. Currently, the tool is adapted to 29 mainstream open source models, and can simultaneously identify unsafe content and the risk of possible information fabrication during the generation phase.

On September 17, Interface News learned that Ant AI Security Lab opened the SingProbe large model security fence and opened the code, models, and evaluation benchmarks. Currently, the tool is adapted to 29 mainstream open source models, and can simultaneously identify unsafe content and the risk of possible information fabrication during the generation phase.