Researchers have uncovered a new cyber threat known as TypoPrompting, which capitalizes on minor mistakes in typing the URLs of popular AI tools. These errors redirect users to malicious sites that may install malware or alter browser settings. This technique marks a concerning development in cybersecurity.
Discovery of TypoPrompting
In a comprehensive study, researchers identified 186 deceptive domains mimicking 27 AI brands. Notable names include Claude, ChatGPT, DeepSeek, and Midjourney. A small typo, such as typing ‘cloude.ai’ instead of ‘claude.ai’, can lead users through a series of redirects to harmful software installations.
The campaign, named TypoPrompting by Island analysts, transforms typical AI tool navigation into deceptive prompts requiring verification or installation actions. The findings, published on November 12, 2026, by security expert Oleg Zaytsev, highlight the significant risks posed by these seemingly innocuous errors.
Mechanics Behind TypoPrompting
TypoPrompting exploits common spelling errors, such as swapping letters or omitting characters, to direct users to malicious destinations. The system assesses the user’s location and device, determining the type of attack to deploy. Some may encounter harmless pages, while others face severe threats.
These domains have been strategically registered, some as early as July 2023, forming a network that misleads users into downloading harmful software. The case of ‘cloude.ai’ illustrates the complexity of these attacks, which involve multiple redirects and fake verification steps.
Implications and Prevention Measures
The broader implications of TypoPrompting are significant, affecting AI brands and beyond. For instance, variations of Scotiabank and Microsoft Azure URLs have been misused. Mac users often face fake terminal prompts, while others receive malicious archives to execute.
Organizations are advised to provide employees with verified AI tool links and monitor for potential lookalikes. Security teams should scrutinize the entire path from initial domain access to endpoint changes. Bookmarking genuine AI tools can mitigate risks, as real verification processes never require command pasting.
As technology advances, the threat landscape evolves, highlighting the importance of vigilance and proactive security measures. Understanding and countering TypoPrompting can protect users from falling victim to these sophisticated cyber threats.
