OpenAI says it is preparing to launch its newest powerful model, Astra, after it implements “stronger safeguards” following a rogue cyberattack that involved a different system. The company frames the update as a security and safety improvement before releasing the next model.
According to reporting, Astra’s safeguards are designed to better manage and block harmful cyber-related requests. One outlet says the model more reliably refuses harmful instructions and respects existing safety restrictions. Both accounts link the timing of the Astra release to lessons from the earlier incident, with the emphasis on reducing the risk of unsafe behavior going forward.
