WormGPT: Dark AI Tool Used by Hackers for Cyberattacks

About the Author: This article was researched using current cybersecurity reports, verified sources from security firms including SlashNext, Netenrich, and Cato CTRL, and authoritative information from organizations including Europol and the National Security Agency. The information reflects the state of dark AI threats as of late 2025.

The rise of artificial intelligence has brought incredible benefits to businesses and individuals worldwide. However, the same technology that powers helpful tools like ChatGPT has spawned a dangerous counterpart designed specifically for malicious purposes. WormGPT represents a troubling evolution in cybercrime, where AI becomes a weapon in the hands of hackers. As we explore AI trends shaping social media and technology in 2026, it's crucial to understand both the opportunities and threats that AI presents.

What Is WormGPT?

WormGPT is an AI-powered chatbot created explicitly for cybercriminal activities, functioning as what security researchers call a "blackhat alternative" to legitimate AI tools. Unlike ChatGPT or other mainstream language models that include built-in ethical safeguards and content moderation, WormGPT operates without any moral boundaries or restrictions.

Security researchers from SlashNext first discovered WormGPT being advertised on dark web hacker forums in July 2023. The tool was built on the GPT-J language model, an open-source framework developed in early 2021, but was trained specifically on malware-related data to optimize it for criminal applications.

The creator, later identified as Rafael Morais, claimed the tool was designed to be "neutral and uncensored." However, its functionality tells a different story. WormGPT enables users to generate phishing emails, create malicious code, and conduct sophisticated cyberattacks without the refusals typical of legitimate AI assistants.

How WormGPT Differs From Legitimate AI Tools

The fundamental difference between WormGPT and tools like ChatGPT lies in their design philosophy and operational constraints:

Ethical Guardrails: Mainstream AI platforms incorporate multiple layers of content moderation to prevent misuse. They refuse requests for malicious code, phishing templates, or harmful content. WormGPT deliberately removes these safety features, responding to any request regardless of its illegal or unethical nature.

Training Data: While legitimate AI models are trained on diverse, publicly available information with strict filtering, WormGPT was allegedly trained on datasets focused specifically on malware, hacking techniques, and cybercriminal tactics. This specialized training makes it particularly effective at generating malicious content.

Accessibility: ChatGPT and similar tools are publicly available through legitimate channels with transparent pricing and terms of service. WormGPT is sold exclusively on dark web forums and through encrypted communication channels, with pricing that initially reached €500 (approximately $540) per month for access, with some users paying thousands for private installations.

Purpose and Intent: Legitimate AI tools aim to assist users with productivity, creativity, and problem-solving within ethical boundaries. These tools are increasingly becoming part of AI-augmented work environments, transforming how businesses operate. In contrast, WormGPT was designed from the ground up to facilitate illegal activities, with marketing materials explicitly highlighting its usefulness for conducting cyberattacks.

The Real-World Threat: Business Email Compromise Attacks

One of the most concerning applications of WormGPT is its use in Business Email Compromise attacks. BEC is a sophisticated form of cybercrime where attackers impersonate company executives or trusted vendors to trick employees into transferring money or disclosing sensitive information.

Former black hat hacker Daniel Kelley, now working with cybersecurity firm SlashNext, conducted tests using WormGPT to assess its threat potential. His findings were described as "unsettling." The AI produced emails that were not only remarkably persuasive but also strategically sophisticated, demonstrating capabilities that could make even novice cybercriminals dangerous.

Traditional BEC attacks required significant skill to craft convincing messages that mimicked executive communication styles, included appropriate contextual details, and created a sense of urgency without raising suspicion. WormGPT automates this entire process, generating highly credible impersonation emails in seconds. The tool can dynamically create tailored phishing lures targeting specific financial workflows and support high-volume targeting at scale.

According to cybersecurity experts, WormGPT's BEC capabilities include the ability to analyze publicly available information about companies and individuals, craft personalized messages that reference specific projects or relationships, and maintain consistent communication threads that build trust before making fraudulent requests.

The Broader Dark AI Ecosystem

WormGPT is not an isolated phenomenon but part of a growing ecosystem of malicious AI tools emerging in the cyber underground. Several similar platforms have appeared since WormGPT gained notoriety:

FraudGPT positions itself as an all-in-one solution for cybercriminals, offering capabilities beyond just phishing emails. For subscription fees ranging from $200 per month to $1,700 per year, users can access features for writing phishing emails, creating malware and hacking tools, discovering system vulnerabilities, finding compromised credentials, and accessing advice on various hacking techniques and cybercrime methodologies. By late 2023, FraudGPT claimed over 3,000 confirmed sales, indicating significant adoption within criminal communities.

DarkBERT and DarkBART represent the next evolution of these tools, with developers claiming they were trained on data scraped from the entire dark web. These systems allegedly offer enhanced capabilities including integration with image analysis tools and access to comprehensive underground knowledge bases. While legitimate AI picture generators create art and helpful visuals, these dark tools manipulate images for fraudulent purposes.

WormGPT Variants have proliferated following increased attention to the original tool. Newer versions like xzin0vich-WormGPT and keanu-WormGPT are actively promoted on forums like BreachForums. Recent research from Cato CTRL reveals that some of these variants are powered not by obscure models but by hijacked versions of high-profile, legitimate AI systems that have been stripped of their safety features.

Security researchers note that these tools represent a concerning trend toward the democratization of cybercrime. Where sophisticated attacks once required years of technical expertise, these AI-powered tools lower the barrier to entry dramatically, enabling even relatively unskilled individuals to conduct complex cyberattacks. While legitimate chatbot platforms help beginners learn AI technology, dark AI tools exploit similar interfaces for criminal purposes.

Why Dark AI Tools Are So Dangerous

The emergence of WormGPT and similar tools fundamentally changes the threat landscape in several critical ways:

Scale and Speed: Traditional cybercriminals might send dozens or hundreds of phishing emails daily, each requiring manual crafting and customization. With AI assistance, that number can scale to thousands or tens of thousands with minimal additional effort. The speed of attack generation far exceeds what human operators could achieve.

Sophistication Without Expertise: Previously, conducting convincing BEC attacks or creating effective malware required significant technical knowledge and social engineering skills. Dark AI tools encapsulate this expertise within automated systems, allowing novice criminals to execute attacks that would have required expert-level capabilities just a few years ago.

Personalization at Scale: One of the most powerful aspects of these tools is their ability to process vast amounts of publicly available information about targets and generate highly personalized attacks. They can analyze social media profiles, company websites, press releases, and other data sources to create messages that include specific, credible details about the target's work, relationships, and circumstances.

Evasion of Traditional Defenses: Many email security systems rely on pattern recognition to identify phishing attempts, looking for common grammatical errors, suspicious phrasing, or known malicious patterns. AI-generated messages often appear flawless in their grammar and structure, using natural language that closely mimics legitimate communication, making them much harder for automated systems to flag.

Continuous Evolution: Unlike static malware or fixed attack templates, AI-powered tools can adapt their output based on feedback and changing circumstances. If certain approaches become less effective, the underlying models can adjust their strategies, creating an arms race between attackers and defenders.

Europol highlighted these concerns in a 2023 report, warning that dark large language models could become future criminal business models, fundamentally transforming the cybercrime landscape. The organization emphasized the need for law enforcement and cybersecurity professionals to develop new strategies to counter these AI-enabled threats.

How Criminals Access and Use WormGPT

The distribution and usage of WormGPT follows patterns typical of dark web criminal services, but with some unique characteristics that make it particularly accessible to a broad range of actors.

WormGPT and similar tools are primarily advertised through specialized dark web forums that cater to cybercriminals. These platforms operate using Tor and other anonymization technologies, making it difficult for law enforcement to track users or shut down services. Access typically requires cryptocurrency payments, usually Bitcoin or Monero, which provide additional layers of anonymity.

The subscription model employed by these tools mirrors legitimate software-as-a-service businesses, making them easy to access for those with criminal intent. Users simply pay their monthly or annual fees and receive credentials to access the AI interface through encrypted web portals or dedicated applications.

Some versions of WormGPT are also promoted through encrypted messaging platforms like Telegram, where vendors establish channels to advertise their services, share updates, and provide customer support to their criminal clients. This migration from purely dark web forums to more accessible platforms indicates an expansion of the potential user base.

The original WormGPT gained over 200 subscribers within its first few months of operation, generating significant revenue for its operators. While the original tool was eventually shut down following increased scrutiny and law enforcement attention, numerous variants have emerged to fill the void, suggesting strong demand within criminal communities.

Real-World Consequences: Case Studies and Examples

While many WormGPT-specific attacks remain unreported due to the stigma and potential liability associated with cybersecurity breaches, the broader impact of AI-powered phishing and BEC attacks provides clear evidence of the threat's severity.

Financial losses from BEC attacks have reached staggering levels. These attacks increasingly leverage AI-generated content to improve their success rates. Organizations across all sectors have fallen victim to these sophisticated schemes, with some individual incidents resulting in losses exceeding millions of dollars.

The technology enabling tools like WormGPT has also been used in deepfake fraud cases. In one notable incident, fraudsters used AI-generated images and video to impersonate actor Brad Pitt, convincing a woman in France that she was in a romantic relationship with the celebrity and ultimately defrauding her of over $800,000. While not directly attributed to WormGPT, this case demonstrates the devastating potential of AI-powered social engineering.

Healthcare organizations, financial institutions, and government agencies represent particularly attractive targets for these attacks. The sensitive nature of their data and the high-value transactions they regularly process make them prime candidates for AI-enhanced BEC schemes. Several cybersecurity firms have reported observing campaigns that bear the hallmarks of AI-generated content, including unusually sophisticated personalization and rapid iteration of attack approaches.

Small and medium-sized businesses face particular vulnerability, as they often lack the sophisticated email security systems and employee training programs that larger organizations employ. A single successful BEC attack can devastate these companies financially, and the increasing accessibility of tools like WormGPT means attackers can target hundreds or thousands of smaller organizations simultaneously.

How to Protect Yourself and Your Organization

Despite the sophisticated nature of AI-powered threats, organizations and individuals can take concrete steps to protect themselves from attacks leveraging tools like WormGPT.

Implement Robust Email Security: Deploy advanced email filtering solutions that use machine learning to detect anomalies in communication patterns. Look for systems that analyze not just content but also metadata, sender behavior, and contextual factors. Enable DMARC, SPF, and DKIM authentication protocols to prevent email spoofing and verify sender legitimacy.

Establish Verification Procedures: Create and enforce strict verification processes for any financial transactions or sensitive information requests, especially those received via email. Implement multi-channel verification where any email request for money transfer or credential sharing must be confirmed through a secondary channel like phone calls or in-person communication. Maintain a list of verified contact information for executives and key vendors, ensuring employees confirm requests using these trusted channels rather than contact information provided in suspicious emails.

Conduct Regular Security Awareness Training: Educate employees about the evolving nature of phishing attacks, including the increasing sophistication of AI-generated content. Understanding how to detect AI-generated content can help employees identify suspicious communications. Provide specific examples of BEC attacks and teach employees to recognize warning signs like unusual urgency, requests that bypass normal procedures, or slight anomalies in sender addresses. Conduct simulated phishing exercises using realistic scenarios to test and improve employee vigilance.

Implement Multi-Factor Authentication: Require MFA for all business email accounts and critical systems. Even if credentials are compromised through phishing, MFA provides an additional barrier that can prevent unauthorized access. Use hardware security keys or authenticator apps rather than SMS-based verification when possible, as these methods offer stronger protection against sophisticated attacks.

Maintain Minimal Public Information: Review what information about your organization and employees is publicly accessible on websites, social media, and other platforms. Cybercriminals use this data to craft convincing phishing messages. While automating social media can boost your marketing, be mindful of what organizational details you share publicly. Consider limiting the details available about organizational structure, ongoing projects, and individual responsibilities. Provide training on personal privacy practices, encouraging employees to minimize their digital footprint.

Deploy AI-Powered Defense Systems: Fight fire with fire by implementing security solutions that use artificial intelligence to detect and block AI-generated attacks. Just as businesses use AI generator tools for legitimate business and content creation, security teams can leverage AI to identify threats. These systems can identify subtle patterns that suggest automated content generation or analyze communication for indicators of social engineering attempts.

Create an Incident Response Plan: Develop clear procedures for responding to suspected phishing attacks or security breaches. Ensure employees know whom to contact if they receive suspicious communications or accidentally interact with potential phishing content. Quick response can minimize damage from successful attacks.

Monitor Financial Systems Closely: Implement real-time monitoring of financial transactions and establish alerts for unusual activities. Create approval workflows that require multiple levels of authorization for transactions above certain thresholds. Regularly review and reconcile accounts to quickly identify unauthorized transfers.

The Future of Dark AI and Cybersecurity

The emergence of WormGPT represents just the beginning of AI-powered cybercrime. Security experts predict that these tools will continue evolving, becoming more sophisticated and more accessible to criminal actors. To stay ahead of both positive and negative developments, follow our tech trends coverage for the latest updates on AI and cybersecurity.

Several concerning trends are already emerging. AI models are being trained on increasingly specialized datasets, creating tools optimized for specific types of attacks like ransomware deployment, credential theft, or social media manipulation. The integration of multimodal AI capabilities means future tools may combine text generation with voice synthesis and image manipulation, enabling even more convincing deepfake attacks. Automation is advancing to the point where entire attack chains can operate with minimal human oversight, from initial reconnaissance through exploitation and data exfiltration.

On the defensive side, cybersecurity companies are racing to develop AI-powered protection systems that can detect and counter these threats. Machine learning models trained on attack patterns can identify AI-generated content with increasing accuracy. Behavioral analysis systems can flag unusual communication patterns that might indicate BEC attempts. However, this creates an ongoing arms race between attackers developing more sophisticated tools and defenders creating better detection systems.

Regulatory responses are beginning to emerge as well. In June 2025, the National Security Agency issued warnings about nation-state actors leveraging AI capabilities for cyber operations. Governments worldwide are considering legislation to address the development and deployment of AI tools designed explicitly for malicious purposes. However, the international nature of cybercrime and the decentralized structure of the dark web make enforcement challenging.

The cybersecurity community emphasizes that technology alone cannot solve this problem. Human vigilance, organizational culture, and ongoing education remain critical components of effective defense. As AI tools make attacks more convincing and harder to detect automatically, the ability of individuals to recognize and respond appropriately to suspicious communications becomes even more important.

Conclusion

WormGPT exemplifies a disturbing trend in cybercrime where powerful AI technology is deliberately weaponized against individuals and organizations. By removing the ethical guardrails that govern legitimate AI tools, creators of platforms like WormGPT have dramatically lowered the barrier to entry for sophisticated cyberattacks.

The threat is not theoretical. These tools are actively being used to conduct business email compromise attacks, generate malware, and facilitate various forms of cybercrime. The subscription model and active marketing on dark web forums indicate a thriving market for these malicious services.

However, awareness and preparedness can significantly reduce the risk. Organizations that implement comprehensive security measures, train employees to recognize sophisticated attacks, and establish verification procedures for sensitive requests can protect themselves effectively even against AI-enhanced threats.

As artificial intelligence continues to evolve, both its beneficial applications and its potential for misuse will expand. The cybersecurity community must remain vigilant, adapting defense strategies to counter emerging threats while advocating for responsible AI development that includes robust safeguards against malicious use.

Understanding threats like WormGPT is the first step toward building resilience against them. By staying informed about the tactics and capabilities of modern cybercriminals, individuals and organizations can make informed decisions about their security posture and take proactive steps to protect their assets and information in an increasingly complex threat landscape. For those looking to enhance their digital literacy, understanding how search engines work and basic SEO principles can also help you recognize suspicious websites and protect your online presence.