Island, a cybersecurity firm, reported on November 12, 2026 that 186 look‑alike domains impersonating popular AI tools are delivering malicious payloads, a pattern the firm calls TypoPrompting (Island.io).
The domains cover 27 AI brands—Anthropic Claude, OpenAI ChatGPT, Google Gemini, Microsoft Copilot, DeepSeek, Perplexance, Midjourney, ElevenLabs, and Cursor—and were registered as early as July 13, 2023 (The Hacker News).
One delivery route presents a fake Cloudflare verification page that triggers a hidden PowerShell command fetched from a decimal‑encoded IP address; this ClickFix lure attempts to run clipboard commands on Windows or shell commands on macOS.
A second route leads to “Security Check” pages that install Net Guard, a Microsoft Store‑signed application whose package hash matches the Store’s but contains browser‑hijacking code capable of changing the visitor’s default browser and executing remote scripts via DevTools.
The third pathway redirects users to a family of search‑hijacking extensions; 30 variants were built from the same kit, 13 have been removed, and the remaining 17 list a combined 972,000 declared users on the Chrome Web Store.
Island notes that the ClickFix gate script also serves a broader set of 211 entry domains, extending the abuse to look‑alikes of banks, insurers, and enterprise software such as Scotiabank, Charles Schwab, Verizon, ServiceNow, and Microsoft Azure.
The campaign relies on parked domains that initially appear empty—a technique highlighted by Infoblox in December 2025 as a way to monetize traffic through direct‑search advertising networks—though Island’s report does not confirm any specific ad platform involvement.
Defenders should block the enumerated typo domains, monitor DNS queries for sudden redirects, and deploy endpoint protection that can detect the PowerShell and Net Guard payloads; developers, meanwhile, need to verify URLs before downloading AI‑related tools.
The TypoPrompting campaign shows how simple typographical errors can be weaponized at scale, leveraging both legitimate advertising infrastructure and signed Microsoft Store packages to increase trust. The use of a signed app suggests the attackers are targeting platforms where users are less likely to question legitimacy, while the large user base of the search‑hijacking extensions indicates a profitable monetization layer. Continued monitoring of typo‑squatted domains and rapid takedown of malicious extensions will be essential to limit exposure.
Spotted an error? See our corrections policy or email editorial@lordofclaude.com.





Discussion & Comments (0)
Join the conversation