Yeah, "slowing down development" isn't going to happen, because China and open models will not slow down.
Also, OpenAI saying they won't go public "until they have a safe agent" is impossible. No model is "safe," and even less so the more complex/training data they get. SO OpenAI isn't going public then? I coubt that.
Anthropic tested Z.ai's open-weight model GLM-5.3 for hacking skills. It builds exploits almost as reliably as Claude Mythos Preview, found new bugs in a browser and wrote an exploit for an old Chrome flaw for $20.40. Anyone can download the weights, and the safeguards are easy to strip out.
English
Español
Français
Italiano
Nederlands
Polski
Português
Русский
Türkçe
Svenska
Chinese
Magyar