23 Hrs
Source:
البلاد
US AI models surpass Kimi K3 in security
Chinese artificial intelligence models, particularly "Kimi K3," have been evaluated by the UK's Cybersecurity Institute and the US Center for AI Standards and Innovation. The results showed that the model’s performance in cybersecurity capabilities was weak compared to leading American models. "Kimi K3" achieved a score of 32.2% in developing exploits for cybersecurity vulnerabilities, versus 76.2% for the American models. It also failed to execute millions of random lines of code, which is concerning, whereas American models succeeded in 20 out of 41 attempts. These findings indicate that China still lags behind the United States in AI cybersecurity, which eases concerns about its dominance in cyberattack capabilities and diminishes overly optimistic assumptions about China’s progress in this field.
Size: 120%Text Size
Notice: This Is an AI-Generated Summary
Comments (0)