Regulation will restrict public access to open-source AI models that could be used for personal cybersecurity