Cloudflare has released a tortuous new trap for data-hungry AI robots that ignore website permissions – the “AI Maze”.
The AI Maze attempts to actively undermine the AI robot by providing realistic-looking pages filled with irrelevant information and hidden links that go deeper into AI-generated nonsense rabbit holes.
“When we discover unauthorized crawls, rather than blocking requests, we will link to a series of AI-generated pages that convincingly attract crawlers to travel through them,” he said. Cloudflare reveals.
“But, despite what it looks like, these contents aren’t actually the content of the website we are protecting.”
This is exactly how the system works:
- It produces compelling fake pages with scientifically accurate but irrelevant content
- The hidden links hidden in these pages lead to more fake content, creating endless loops
- All trap content is still completely invisible to human visitors
- Interacting with robots with these fake pages helps improve detection systems
- Content is pre-generated, not created on demand in search of better performance
- Crawlers waste their resources, not Cloudfares’ resources
Such tools are needed because robots’ internet traffic is shocking.
according to Imperva’s 2024 Threat Research Reportrobots generated 49.6% of network traffic last year, and malicious robots accounted for 32% of the total.
AI Crawlers have over 50 billion requests bombard Cloudfare’s network every day (which accounts for nearly 1% of the network traffic it processes) wasted resources in the process.
These figures make the credibility of many people consider “dead Internet theory” – Internet conspiracy claims that most online content and interactions are artificially generated.
CloudFlare attempts to support customers in a cat-drive game between website owners and AI companies. Traps are still completely invisible to human tourists, so they should not accidentally stumble upon the maze.
As CloudFare describes: “No real person goes deep into four links and becomes a maze of nonsense generated by AI. Any visitors who really are likely to be robots, so this gives us a brand new tool to identify and fingerprint bad robots, which we add to our list of known bad actors.”
