A new AI tool has appeared on the dark web and quickly attracted the attention of security experts, and not for the best of reasons.
This is an artificial intelligence service called DIG AI, which has no built-in limitations. This bot is already actively used in fraudulent schemes, malware development, the dissemination of extremist material, and the creation of content related to the sexual exploitation of minors.
Researchers at Re Security first detected traces of DIG AI on September 29, 2025. Almost immediately after its launch, the service’s administrator began actively promoting it on a darknet forum, boasting about its workload: in the first 24 hours, the system had processed around ten thousand requests.
Unlike previous AI tools for organized crime like FraudGPT or WormGPT, which were sold on a subscription basis, DIG AI is designed differently. It doesn’t require registration, payment, or even an account: it simply requires access via the Tor network. Furthermore, the creator states that the service is deployed on its own infrastructure and does not rely on third-party clouds, further improving its resilience to blockades.
Resecurity conducted a series of tests and concluded that the bot readily answers questions related to the production of explosives, drugs, other prohibited substances , financial fraud, and other topics prohibited by international law. During the experiments, the system generated functional malicious scripts, including code for installing backdoors and other types of malware . Analysts believe the results are quite suitable for practical use.
DIG AI’s capabilities are attracting particular attention in the cybersecurity community, particularly in relation to its ability to process pornographic content. The tool was capable of both creating entirely synthetic materials and modifying images of real minors, transforming harmless photographs into illegal material. Experts call this aspect one of the most alarming.
Despite all its potential, the service still has limitations. Some operations take several minutes to complete, indicating limited computing resources. However, this problem can be easily addressed, for example by introducing paid access and scaling the hardware based on demand.
Meanwhile, several TOR platforms are displaying DIG AI advertising banners associated with drug trafficking and the resale of compromised payment data. This selection of platforms accurately describes the service’s developer’s target audience. The administrator, who uses the pseudonym Pitch , claims that one of the three available templates is based on ChatGPT Turbo.
Between 2024 and 2025, mentions and actual use of malicious AI tools on underground forums tripled. Cybercriminals are increasingly mastering large-scale language models , and the emergence of new systems is only accelerating this process.
Obviously, these technologies could lead to a significant increase in threats as early as 2026. This concern is accentuated by the major international events planned for that period, including the Milan Winter Olympics and the FIFA World Cup .
Analysts believe that criminal AI systems are lowering the barriers to entry for cybercrime by automating and amplifying attacks, thus expanding the pool of potential attackers.
Follow us on Google News to receive daily updates on cybersecurity. Contact us if you would like to report news, insights or content for publication.
