This AI autonomous offense-and-defense round is extremely hardcore—GLM 5.2 local forensics is the real security baseline.

Crypto界网消息，on-chain analyst AI姨 said that this real AI vs AI attack-and-defense battle was driven end-to-end by an autonomous AI agent system, and the team mainly relied on its own AI to detect and analyze. The actions were carried out by an autonomous agent framework; the specific LLM used is still unclear. It executed thousands of independent operations, and the command-and-control servers self-migrated and deployed on public services, fitting the industry-predicted “agentic attacker” scenario. In log analysis, the team attempted to use cutting-edge models behind commercial APIs, but could not submit real attack commands due to security guardrails intercepting them. In the end, they switched to GLM 5.2 to perform forensics analysis on their own infrastructure, ensuring that attacker data never left the environment.