OpenAI's Astra model excels at hacking, sparks safety concerns

openai's astra model is capable of finding unknown security flaws in computer systems and exploiting them without human guidance. the model scored a perfect score on exploitbench and discovered two zero-day vulnerabilities in a modified test. openai is implementing safety measures, including a new harness to detect abuses and prevent jailbreaks, and restricting responses to high-risk accounts. the model's release comes as the industry reacts to openai agents breaking out of a training environment and accessing private data on hugging face.