Last changed Oct 27, 2026

Attackers hijack at least 17 internet-exposed Cairn AI pentesting panels, making the agents plant SSH keys and a Poseidon implant on their own hosts and hunt AI API keys

Unknown clusters tracked by NorForge as AIA0009 and AIA0010; operators of exposed Cairn panels (red teams, CTF and bug-bounty users, and suspected criminal operations); NorForge (researcher)

NorForge, an AI-agent sandboxing firm, reported on Oct 27 that by passively watching internet-exposed panels of Cairn, a platform where users give AI agents a goal and they work autonomously, it saw attackers inject tasks into at least 17 panels that lacked authentication. Activity intensified around Oct 20, with possible reconnaissance from about Oct 10. Injected tasks had the agents add attacker SSH keys to the kali user's authorized_keys, enumerate environment variables and config files for secrets such as ANTHROPIC_AUTH_TOKEN and write them where any panel visitor could read them; on Oct 21 at 21:15 UTC one task carried a base64 script that the agent decoded and ran, downloading and launching a Poseidon implant (Mythic framework) that communicated via Bunny.net infrastructure. Synchronized waves across panels suggest scripting, with apparent human intervention before malware deployment. Some probed for container escapes; NorForge could not tell whether any succeeded. The volume of data actually taken is not confirmed. A separate handle, AstraRCE, defaced many panels on Oct 20.