New TypoPrompting Attack Targets AI Tools

A new 'TypoPrompting' campaign uses 186 fake AI domains to spread browser-hijacking malware and redirect user searches.

Security researchers have identified a new threat campaign named "TypoPrompting" that leverages at least 186 fraudulent domains designed to look like popular artificial intelligence tools. According to a report from The Hacker News based on findings by the firm Island, the campaign deceives users who make a small spelling mistake when typing a URL. Instead of reaching their intended AI service, they are redirected to malicious sites. These sites distribute a variety of unwanted software, including browser-hijacking malware, search hijackers, and deceptive ad-serving programs. This tactic specifically targets the rapidly growing user base of AI applications, turning a simple typo into a significant security risk for individuals and organizations alike.

The attack vector is a refined version of a classic cybersecurity threat known as typosquatting. The process begins when a user, often a developer or tech professional, attempts to navigate to a well-known AI platform but misspells its domain name. The fake website they land on is engineered to look legitimate, often copying the branding and user interface of the real service. It then presents a deceptive pop-up or prompt, such as a fake CAPTCHA, a browser update notification, or a request to install a "helper" extension. If the user interacts with this prompt, they unknowingly authorize the installation of a malicious browser extension or the execution of a script. Once installed, this malware can take control of the browser's default search engine, redirecting all queries through a third-party service to generate fraudulent ad revenue or display malicious search results.

This TypoPrompting campaign is not an isolated incident but part of a broader trend where attackers quickly exploit emerging technology ecosystems. We have seen similar patterns with the rise of cryptocurrency, where fake exchange sites and wallet apps proliferated, and with the growth of popular social media platforms. The current "AI gold rush" creates a perfect storm for such attacks. Employees are often encouraged to experiment with new AI tools to boost productivity, leading to a high volume of traffic to new and unfamiliar domains. Threat actors capitalize on this enthusiasm and the lack of established security muscle memory around these new services. By targeting developers and IT staff—the primary users of many cutting-edge AI tools—attackers aim for high-value targets who often possess elevated permissions and access to sensitive corporate infrastructure.

For organizations, the emergence of TypoPrompting underscores the need for proactive security measures that go beyond traditional antivirus software. The primary defense is user awareness and education. Teams should be trained to spot the signs of a typosquatting site and be cautious of unexpected prompts or extension installation requests, even on seemingly familiar websites. A simple but effective best practice is to use bookmarks for frequently accessed services rather than manually typing URLs. From a technical standpoint, security teams should consider deploying web filtering solutions to block known malicious domains and using browser management policies to restrict the installation of unverified extensions. As the AI landscape continues to expand, security leaders must anticipate that attackers will continue to mimic popular tools, making ongoing vigilance and adaptive security postures essential.

This campaign highlights a growing threat vector targeting developers and technical teams who are early adopters of AI tools. A simple typo when accessing a trusted service can lead to compromised browsers, data exfiltration, and the introduction of malware into secure development environments.

Employees falling for these typosquatting schemes can expose corporate networks to significant security risks, including data breaches and credential theft. This increases the burden on IT and security teams, requiring enhanced monitoring, user training, and potentially costly incident response efforts.