Threat Actors Phish Credentials Using Fake ChatGPT, Claude, DeepSeek Brands

Key Takeaways

• Cybercriminals are exploiting the popularity of AI platforms like ChatGPT, Claude, and DeepSeek to launch sophisticated phishing and malware campaigns.
• These attacks leverage social engineering, impersonating official communications and services to trick users into divulging credentials, credit card details, and authentication tokens.
• Attack chains are increasingly complex, routing victims through legitimate services such as URL shorteners and CRM tools to evade detection.
• Thousands of organizations globally have been targeted, with some campaigns deploying infostealers like Vidar.

Cybersecurity analysts have uncovered a new wave of attacks where threat actors are leveraging the prominent branding of leading artificial intelligence platforms, including ChatGPT, Claude, and DeepSeek, to execute elaborate phishing and malware distribution schemes. These campaigns are meticulously crafted to deceive users into surrendering sensitive information such as login credentials, credit card numbers, and authentication tokens.

The rapid integration of AI tools into daily routines has created a fertile environment for malicious actors. As millions adopt AI assistants, many users are still learning to distinguish authentic communications from fraudulent attempts, presenting a prime opportunity for exploitation.

Attackers capitalize on this by designing fake web pages or emails that mimic trusted AI platforms. A significant number of users, unaware of the deception, proceed to click on malicious links or download compromised files.

Microsoft Threat Intelligence analysts documented several of these campaigns that transpired in early 2026. Microsoft clarified in a report shared with Cyber Security News (CSN) that these operations do not signify any breaches of the AI services themselves. Instead, they are purely social engineering endeavors that exploit established brand trust to coax users into clicking malicious links, opening infected PDFs, or downloading malware.

A notable characteristic of these attacks is their enhanced stealth. Threat actors are routing victims through legitimate, trusted online services before ultimately redirecting them to malicious destinations. This layered approach complicates detection efforts.

Services such as URL shorteners, customer relationship management (CRM) platforms, and even GitHub are integrated into the attack chain to mask the true origin and nature of the malicious content. By the time victims discern the fraudulent nature of the interaction, their sensitive information may have already been compromised. The repercussions of these attacks are significant, with thousands of organizations across numerous countries targeted, leading to losses of credit card data, account access, and authentication tokens that provide direct entry points into corporate systems.

Threat Actors Abuse ChatGPT, Claude, and DeepSeek Brands

One illustrative example is a ChatGPT-themed campaign identified on May 5, 2026. This operation involved approximately 4,500 emails sent to targets in South Africa. The emails falsely warned recipients that their ChatGPT Plus subscription was at risk of downgrade unless they updated their payment method within seven days. These emails featured the authentic ChatGPT logo and a seemingly legitimate “update” button.

Clicking this button did not immediately lead to a malicious site. Instead, victims were routed through a complex chain involving a CRM service, an Amazon tracking domain, and a URL shortener before finally landing on a compromised website. This site hosted a fake payment page within a subfolder, complete with a deceptive CAPTCHA to deter automated scanners. The page then proceeded to collect personal details and full credit card information over a two-step process.

Another campaign, leveraging the Claude brand, was active from April 20 to 22, 2026, impacting over 2,000 organizations in the United States, the United Kingdom, and India. Emails in this campaign alleged that the recipient’s account had violated usage policies. An attached PDF, titled “Fill and Sign Claude Appeal Form.pdf,” directed users to an attacker-controlled domain. Following a series of fake verification screens, victims were ultimately redirected to what appeared to be a Microsoft sign-in page, specifically designed to steal access tokens.

Fake DeepSeek Installer and Malvertising Drop Vidar

In April 2026, threat actors swiftly capitalized on the preview release of DeepSeek’s V4 model. Within 45 minutes of the announcement, a fraudulent GitHub organization named DeepSeek-V4 was established. This fake repository incorporated stolen branding, legitimate benchmark data, and search-optimized tags to achieve high rankings in both traditional and AI-driven search results. Users who downloaded archives from this repository inadvertently received a loader that silently installed the Vidar infostealer onto their devices.

Concurrently, a separate malvertising campaign, attributed to Storm-3075, promoted a bogus product called “Awesome AI Windows Plugin” via free movie streaming websites. The download provided was a fraudulently code-signed executable, linked to Fox Tempest, a group known for operating a malware-signing service utilized by various criminal actors. Upon launching the file and clicking a “Continue” prompt, a Python downloader discreetly retrieved the Vidar infostealer from an attacker-controlled server.

What You Should Do

• Enable multi-factor authentication (MFA) on all accounts to add an essential layer of security.
• Exercise extreme caution with unsolicited emails and messages; avoid clicking suspicious links or downloading attachments from unknown sources.
• Always verify communications from AI platforms by directly visiting their official websites rather than relying on links provided in emails.
• Organizations should implement robust email link scanning tools and solutions capable of detecting and blocking phishing pages before they reach end-users.
• Regularly update operating systems, applications, and security software to protect against known vulnerabilities.
• Educate users on the tactics of social engineering and the importance of recognizing phishing attempts.

Disclaimer: HackersRadar reports on cybersecurity threats and incidents for informational and awareness purposes only. We do not engage in hacking activities, data exfiltration, or the hosting or distribution of stolen or leaked information. All content is based on publicly available sources.