Cloudflare推出智能体行为评估新机制
Unveiling good and bad behaviors on the Agentic Internet
Cloudflare出了新系统，能根据行为判断是好人还是坏机器人，还能用他们的模拟工具测测自己的光标像不像真人。
Cloudflare将机器人缓解从单次风险评分转向持续信任评估，通过BotBase和Precursor系统分析代理在互联网上的行为模式。
新系统可区分善意与恶意的自动化行为，并公开了Precursor Trace模拟工具，让用户测试自己的光标移动轨迹被判定为人类还是机器人的概率。该方案针对基于大模型的智能体日益增多的现状，强化网站防护能力。
Unveiling good and bad behaviors on the Agentic Internet
Cloudflare is shifting bot mitigation from point-in-time Risk assessment to continuous Trust evaluation. Learn how new good and bad behaviors from bots and agents are assessed by our systems, including BotBase and Precursor — and try out our Precursor Trace simulation to see how your own cursor movements would be assessed as human or bot.
This report is categorized under AITOP's "Products" desk. Key entities and technical topics under continuous tracking include Cloudflare, BotBase, Precursor, 智能体.
Primary source verification: published directly by Cloudflare Blog with first-party authority.
