As ChatGPT, an artificial intelligence tool, gains in popularity, more and more nefarious use cases have started to appear online. Recently, it has been reported that cybercriminals are circumventing bot restrictions in order to develop tools for hacking and cyber fraud.
According to a report from cyber security firm Check Point Research (CPR), the cybercriminal community has already expressed strong interest in this latest trend to develop malicious code.
“Several major underground hacking communities show that there are already first instances of cybercriminals using OpenAI to develop malicious tools,” the report states adding that cybercriminals using OpenAI have no knowledge of code development. “It’s only a matter of time until more sophisticated threat actors enhance the way they use AI-based tools for bad.”
Can the ChatGPT results be tampered with? The bot responds
ChatGPT is a version of the GPT (Generative Pre-trained Transformer) model, which is trained on a vast corpus of text, including books, papers, and websites. The model is trained by feeding it a collection of words, which it then uses to predict the next word in the series. By spotting patterns and connections between words and sentences in the training dataset, the machine learns to produce coherent and dynamic text. The model can be fine-tuned for certain tasks, such language translation or question answering, after it has been trained.
When we asked ChatGPT if it could be manipulated, the AI bot admitted that its output could be influenced.
ChatGPT answers, “One way this can be done is by providing the model with biased or misleading input data during fine-tuning. If the fine-tuning data contains biased information, the model may learn and reproduce that bias in its output.”
It adds that it is possible to negatively affect output by providing the model with a specific prompt or seed text that guides the model’s output in a certain direction.
