OpenAI proposes to phase out new models to protect against cybersecurity risks

According to an informed source, OpenAI is refining a model with advanced cybersecurity capabilities and plans to publish it to only a few companies, similar to the limited Anthropic release of Mythos. Artificial intelligence has reached a critical point, at least in terms of autonomy and hacking. Model manufacturers are now very concerned about the potential damage to their own tools and are reluctant to make them publicly available. On Tuesday, Anthropic stated that it was planned to limit its new Mythos preview model to a selected group of technology and cybersecurity companies because of concerns about its advanced hacker capabilities. It was the first AI company to adopt this approach to the new model. According to the source, OpenAI now plans to do the same. (Grunting)