【#Tech24H】According to TechCrunch, OpenAI recently released a major announcement: the cyberattack capabilities of its self-developed next-generation large model Astra have exceeded expectations. Out of cybersecurity considerations, the company has decided to suspend partial research and development work on the project and restrict all testing access. Internal assessments by the R&D team revealed that Astra, which is still under development, has seen drastic improvements in its agent code generation and network intrusion capabilities, crossing critical high-risk cybersecurity boundaries. The model is capable of autonomously locating highly secured systems and launching targeted attacks. Per the company’s mature AI risk control framework, multiple layers of safety guardrails must be immediately implemented for the model. [ By Zhang Liyan ]