Researchers warn that the growing “middle class” of AI models may become a bigger offensive security threat than frontier systems because they are cheaper, increasingly capable, and easier to run repeatedly at scale. New testing showed strong exploitation and vulnerability-finding performance from models including GPT 5.5, GLM-5.2, Grok 4.5, Opus 4.7, and Meta’s Muse Spark 1.1, with agent swarms and live-system interaction proving especially powerful. #GPT5_5 #GLM52 #Grok45 #Opus47 #MuseSpark11 #Mythos #XBOW #Anthropic #OpenAI #Meta #xAI #Zai
Keypoints
- Mid-tier AI models are becoming strategically important in offensive security.
- Cheaper models can outperform frontier systems when run longer and more often.
- GPT 5.5 showed major gains in autonomous web application testing and exploitation.
- Live interaction with a target mattered more than source code access in many tests.
- Coordinating multi-agent swarms found far more vulnerabilities than solo agents.
Read More: https://cyberscoop.com/mid-tier-ai-models-hacking-threat/