INFOMENTAL

#OpenAISaysAstraFindsFlawsAutonomously OpenAI says that its next model finds security flaws that nobody has discovered until now. With the right tools and access, Astra can uncover security vulnerabilities that are unknown to date and develop ways to exploit them,” said Amelia Glaese, OpenAI’s vice president in charge of its security work. This is how the company describes its own model, which has not yet been published. Astra detects more security vulnerabilities than any other publicly available OpenAI model and uses fewer computational resources to do so, the company announced on Monday. It will soon be available to a limited group, though the date has not yet been specified. Using fewer computational resources to achieve this is the part that security teams should pay close attention to. A capability that becomes cheaper gets used more widely, and the cost of running a model has historically been the only factor limiting who could do it $AAPLB $GOOGL.US