1-minute read
What a defender needs to know
Huntress found SectopRAT infections at 29 organizations after victims followed paid search results for Claude Desktop. The ads led first to a public Claude Artifact on the legitimate claude.ai domain. Its download button then redirected visitors to attacker-controlled sites serving a fake ClaudeDesktop.exe.
The file was a legitimate signed application paired with a modified DLL, allowing the attackers to run their code through DLL sideloading. Later stages added persistence, anti-analysis checks and command-and-control data stored through BNB Smart Chain transactions.
Public AI artifacts should be treated as user-authored content, even when they share a domain with a provider’s official pages. Security teams should distinguish those paths in browser policy and inspect redirects that lead to software downloads. Approved AI clients also need an easy, managed installation path so employees are not left to choose among search ads.
If SectopRAT is found, response should cover browser sessions, cloud tokens, saved credentials and developer secrets available on the device. Reimaging removes the malware but does not invalidate data the operator may already have stolen.
Verify that active sessions and exposed tokens are revoked, affected accounts show no unexplained changes and the rebuilt endpoint has no recurring SectopRAT process or connection.