Technological advancements allow the creation of several intelligence tools capable of interpreting data and learning from it without human supervision or interference through self-improvement. These are called Artificial Intelligences.
However, they are no longer new to the market. Many applications, stores and services use them. softwares in your favor. One example of them is DarkBERT, an AI trained on the Dark Web itself!
DarkBERT: the new ChatGPT from the deep web
The famous ChatGPT has raised the bar for AI popularity, so now OpenAI’s platform can be used positively in different areas.
However, the creation of malware has also been facilitated by the development of this technology. Over time, even more intelligences like this will be used for negative purposes and trained in different environments.
That’s what just happened to an AI trained on the Dark Web, DarkBERT.

Created in South Korea, this software is based on an AI architecture known as “RoBERTa”, launched in 2019.
The developers saw a different potential that was not taken advantage of at the time of its implementation, being a poorly trained model and, consequently, not offering maximum efficiency. However, the tool was improved.
How was it created?
The researchers trained the RoBERTa model by crawling the Dark Web through firewall from the network known as Tor. With this, they applied data filtering techniques, such as data pre-processing, and from there, they built a database with information from the Dark Web.
In other words, DarkBERT arises from the use of the Large Language Model (LLM) RoBERTa, fed with the information generated by this bank. In this way, the software can analyze data from websites and content on the deep web, even if written in their own languages, and then use them in a useful way.
How can it be used?
The developed model can be used by cybersecurity police officers, as it can reach deep layers of the webwhere illegal trade occurs en masse. DarkBERT can still continue to improve and be applied in areas that have not yet been explored.
Support our work ❤️
If you enjoyed this article, consider leaving a tip to help us keep publishing great content.

























