[ctx]ctxseccontext security: attacks on the harness and context layer of AI agents, plus exploited-now news, hunting and playbooks.

← Threat Hunting · markdown

vendor research (Island, Nov 2026; reported by The Hacker News Nov 12)

Hunt: 'TypoPrompting', 186 typo domains of 27 AI brands lead to ClickFix PowerShell, a Store-signed browser hijacker and search-hijack extensions

  • Examples: cloude[.]ai, clsude[.]ai, cluade[.]ai, caude[.]ai, chatgpy[.]com, chtgpt[.]com, deepseel[.]com, deeepseek[.]com, runawayml[.]com, perplixity[.]ai, lovabe[.]dev, charactor[.]ai.
  • Redirect chain: xml.securetrackingnode[.]com/click → biomimicrydevtech[.]buzz → random subdomain of blockchainfuturedev[.]buzz (FingerprintJS profile plus MD5 proof-of-work before the lure).
  • Endpoint: explorer.exe → powershell.exe from the Run dialog right after browsing; PowerShell fetching from a URL with a decimal IP; installs of a 'Net Guard' Store app; browsers launched with remote debugging enabled.
  • Prevent: block newly registered and typo domains of the AI tools you allow, pin approved AI URLs as bookmarks/SSO tiles, and restrict extension installs to an allowlist.

Sources: Island: TypoPrompting: From AI Typos to Malware (Nov 2026) · The Hacker News: 186 Fake AI Domains Redirect Users to ClickFix Lures, Browser Malware, and Search Hijackers (Nov 12, 2026)