Cloudflare is misleading bots that collect information for AI with fake pages. The company has created its own artificial intelligence that endlessly creates fake pages with unnecessary information, which the bots waste resources on. AI Labyrinth is available to all Cloudflare customers.
Cloudflare is misleading bots that collect information for AI with fake pages. The company has created its own artificial intelligence that endlessly creates fake pages with unnecessary information, which the bots waste resources on. AI Labyrinth is available to all Cloudflare customers.
Cloudflare has unveiled a new initiative called AI Labyrinth, aimed at combating bots that scan websites without permission to collect data and use it in AI training, The Register reports. Instead of traditionally blocking such bots, Cloudflare proposes using generative artificial intelligence to create a network of pages with artificial content, which forces bots to waste resources on unnecessary information.
This technology automatically detects suspicious bot activity and redirects them to specially crafted pages designed and tested with Cloudflare’s Workers AI to avoid security and performance issues.
The goal is to increase the cost of crawling for bot operators, forcing them to spend more resources processing unnecessary content. Cloudflare also sees this tactic as a new way to detect bot activity, as no human will click on multiple links in a maze of meaningless content.
The company came up with the idea for AI Labyrinth when they noticed that nearly one percent of all requests for access to web content they see come from search bots with artificial intelligence. In theory, site owners could block bots or deny them access to the page, but in reality, modern bots are able to bypass almost all such protection methods.
In an interview with Collider, showrunner Tony Gilroy said that he prepared scripts to post online, but then changed his mind.
«I wanted to do this,» he said. «AI is the reason we don’t do it. I mean, unfortunately, it’s too much like X-rays and too easily absorbed. Why help fucking robots more than you can? No real human is going to follow four links deep into a maze of AI-generated nonsense. Any visitor who does is very likely a bot, so this gives us a whole new tool for identifying and fingerprinting bad bots, which we add to our list of known bad actors.»
These kinds of things usually lead to an arms race, and Cloudflare is already thinking about what it needs to do to stay ahead.
«In the future, we will continue to work on making these links more difficult to detect and so that they fit organically into the existing structure of the website in which they are embedded,» the authors write.
