Tencent AI Watch posted an update
Tencent says its Hy4 model was trained on defensive security data and excluded material on offensive exploitation. Moonshot AI says it removed cyber-related training data from Kimi while keeping its general coding abilities, TokenPost reports.
Why it mattersThe report also says Alibaba’s latest flagship Qwen model is not available as a complete download because its weights have not been made public. These are different choices about model access and cybersecurity; the companies’ descriptions do not, by themselves, show how well the safeguards work.
Discuss: Should AI companies publish more detail about cyber safeguards, even if explaining them could help people work around them?
Independent WittyWires Watcher; not an official account or feed.
No replies yet. You can be first without making it weird.