OpenAI gives EU new cyber model access but Anthropic still holding out on Mythos (www.cnbc.com)

OpenAI has announced that it will grant the European Union access to its latest AI model, GPT-5.5-Cyber, a specialized variant focused on enhancing cybersecurity. This limited preview access will be available to vetted cybersecurity teams, including businesses and government authorities across the EU. The move comes in the wake of concerns following the release of Anthropic's model, Mythos, which raised apprehensions about cybersecurity threats. EU officials have expressed appreciation for OpenAI's initiative, stating it will enable closer monitoring of model deployment and tackle security issues more effectively. The significance of OpenAI’s decision lies in its commitment to transparency and collaboration in cyber safety, emphasizing that AI labs should not operate in isolation regarding security matters. By making advanced AI tools accessible to a wider range of cybersecurity defenders, OpenAI aims to bolster collective resilience against potential digital threats. The initiative reflects a proactive approach to ensuring that the latest AI advancements align with European values of shared safety and public security, setting a precedent for partnerships between AI developers and regulatory bodies. In contrast, Anthropic's ongoing negotiations with the EU highlight a more hesitant stance in terms of model access, marking a potentially pivotal moment in the AI regulatory landscape.