Study finds three scanners posing as AI crawlers to hunt server secrets
Start here
A study of ten websites run by the authors between late September and 18 November 2026 identified three scanners — Octopus, Gecko and Tortoise — that impersonate AI crawlers to hunt for secrets on web servers. Octopus rotates among 32 crawler names, including seven undocumented ones such as DeepSeekBot and GrokBot, and rents a fresh Google Cloud machine per site across 15 regions. Gecko sends a Chrome user agent truncated after 60 characters and requests the files where Claude Code, Codex, Cursor and Gemini CLI store keys. The authors say the evidence points to two different operators. On the authors' sites, every request for a secret path in the period received a 404, 403 or 405; the scanners did download public pages and public source maps, which any visitor can do. The authors conclude that blocking by crawler name misses the scanners while hitting real crawlers, and recommend returning 404 for dotfiles and checking request addresses against operators' published ranges.
Generated from reports · updated 1 d ago
Report timeline
Follow the reports to see the event from different sides.
- Hacker News · AIOctopus, Gecko and Tortoise: three scanners hunting AI keys on web servers
A study of ten websites run by the authors from late September to 18 November 2026 found three scanners impersonating AI crawlers to hunt for secrets: Octopus, Gecko and Tortoise. Octopus rotates among 32 crawler names, including seven undocumented ones such as DeepSeekBot and GrokBot, and rents a fresh Google Cloud machine per site across 15 regions, while Gecko sends a Chrome user agent cut after 60 characters and requests files where Claude Code, Codex, Cursor and Gemini CLI store keys. The authors say the evidence points to two different operators, and that blocking by crawler name misses the scanners while hitting real crawlers, so they recommend returning 404 for dotfiles and checking request addresses against operators' published ranges.
Heat for this event
- Current, comparable range
- 5
- Peak, comparable range
- 10November 19 04:00
- Change in 24 hours
- –
The trend compares the same participants over hours observed continuously, so its range can be narrower than the current heat count. Hover or tap the chart for hourly heat; the arrow keys move through it.