| Mobile Web

OpenAI to limit release of new model over security threat concerns like Anthropic

OpenAI plans to give limited early access to a new model with advanced cybersecurity capabilities to only a small number of companies, Axios reported on April 9, citing a source familiar with internal discussions. The approach is similar to Anthropic’s restricted release of Mythos Preview. OpenAI launched an invitation-only “Trusted Access for Cyber” pilot after releasing GPT-5.3-Codex in February, offering model access and $10 million in API credits to participating firms for defensive security work.