Three automated scanners have spent weeks searching websites for secret files that AI tools leave behind, according to a report from the operators of several sites that were targeted. The most active one, which the authors call Octopus, requested about 300 secret files in roughly 10 seconds on each visit. Each request carried the name of a different AI crawler.
What is Octopus, the scanner requesting AI tool secret files?
Octopus starts every visit with a request to a made-up path that ends in the site's own hostname. Two random paths follow, which the authors believe are used to learn how a site answers missing pages. Only then does the scanner ask for files such as .env, cloud credentials and AI gateway settings. The authors counted 32 user agent names in its rotation. Seven of them match no company's documented crawler, including DeepSeekBot and GrokBot. The other 25 copy real crawlers, mostly letter for letter.
How does Octopus avoid being blocked?
Octopus also changes machines constantly. The authors matched every address to Google Cloud's published ranges and found 97 machines in 15 regions between Oct. 3 and Nov. 18. Most lived less than an hour. Of 82 machines in a row, only three used the same region as the one before. A block on one address did nothing, because the next visit came from a new one.
How is Gecko different from Octopus?
Gecko is a different kind of tool. The authors describe it as a .NET program that mostly runs from a single Google Cloud region in Belgium, with some traffic from two rented Windows servers at an offshore hosting company. Its user agent is an old Chrome 128 string that stops after 60 characters, cutting off the words \
Dieser Artikel wurde mit Unterstützung von KI verfasst.
News Factory APP - agentische News für besseres SEO & AEO.
