Open-weight AI models with advanced capabilities and no safeguards are becoming much more accessible. While they can be useful, AI safety experts have concerns.
