Chinese LLMs Broaden the Gap Between Attackers & Defenders
• Cybersecurity firm Semgrep discovered that the Chinese LLM GLM 5.2 outperformed other standard models in testing, achieving a 39% F1 score for identifying true positives and negatives. • The emergence of high-performing models from Chinese firms indicates that frontier AI capabilities are no longer exclusive to US-based companies. • This trend is concerning for cyber-defenders as it broadens the gap between attackers and defenders by making powerful, offensive AI tools more widely available.
darkreading.com