DeepSeek almost matches OpenAI's secret model in cyber hacking tests.
Open-scale AI models have learned to perform computer network hacking tasks almost as efficiently as advanced proprietary designs. This is the conclusion reached by researchers at the company Vals, which evaluates the capabilities of artificial intelligence. In the experiment, China's DeepSeek was only slightly inferior to Daybreak, an OpenAI cyber model available only to a limited circle of trusted partners.