Unveiling good and bad behaviors on the Agentic Internet

The internet's security model is evolving. For decades, the rule was simple: block bots, trust humans. Cloudflare reports that's no longer how the web actually works. These days, sessions constantly shift between human and bot traffic — imagine a user browsing a store, then handing checkout off to an automated assistant. Cloudflare's responding with a new framework based on Trust and Risk: independent but connected values. Trust builds over time through consistent behavior; Risk measures how likely a specific action is harmful. Verified good bots declare themselves honestly and don't abuse earned trust. For catching bad actors, Cloudflare launched Precursor, which monitors how you interact with sites throughout an entire session — cursor movements, acceleration, rhythm, even self-corrections. In just 24 hours, Precursor evaluated over 206 million requests across more than 73 thousand zones. The data confirmed something Cloudflare suspected: suspicious behavior often happens mid-session, and sessions frequently shift from human to automated and back again. The company just launched Precursor Trace, an interactive demo of how these detections work, with more capabilities like Adaptive Intelligence on the way.

Listen to this story

Hear this and more stories in a personalized audio briefing.

Open The Chonkerton