Cloudflare details Trust and Risk tools for hybrid human-agent web traffic

Cloudflare's Web Integrity and Trust team outlined how it separates Risk and Trust when securing sites against hybrid human and agentic traffic on the Cloudflare Blog. The company argues that static bot-versus-human labels no longer fit sessions that shift between people and automated assistants, such as a shopper who hands checkout to a shopping agent. Trust is framed as reputation built over time, while Risk is treated as how harmful a request or action is likely to be. Cloudflare said it is expanding behavior analysis beyond point-in-time checks, including findings tied to Precursor since launch, and plans bot-specific mitigations later in the year. Options described include feeding crawlers LLM-generated page summaries, serving deliberately false content to pollute training data, and queuing legitimate automated traffic such as user-directed shopping agents.