The US government has restricted the release of new AI models from OpenAI and Anthropic due to cybersecurity concerns. The Trump administration has approved a limited release of these models, allowing only trusted partners to access them.
OpenAI’s GPT-5.6 Sol model is being released in stages, with a phased approach to ensure it doesn’t pose any significant risks. The company believes this process will become the default long-term solution.
In contrast, Anthropic has already lifted restrictions on one of its models, Mythos 5, which can now be redeployed to small groups of cyber defenders and infrastructure providers.
Critics argue that government intervention in the AI industry is hindering US companies’ progress. Experts warn that this approach may not address actual risks but rather create unnecessary barriers.
The Trump administration’s actions come after President Donald Trump signed an executive order on AI oversight, which established a framework for vetting national security risks of advanced AI systems before public release.
Source: https://apnews.com/article/trump-ai-openai-gpt56-sol-cybersecurity-mythos-065d5398baac7f16c8265c2cb8ba2baa