Saferai

Open-Weight AI Models Close the Gap, but Safety Still Lags

Open-weight AI models are rapidly approaching frontier performance, but new evaluations show that safety controls still lag behind capability.