However, it raises a question: who should be the culprit? The cybercriminals? The generative AI bots? The firms who develop these bots? Or perhaps the government that fails to come up with proper regulation and accountability?
Generative AI Technology
Generative AI technology is another form of artificial intelligence that aids users in generating texts, images, sounds, and other content from inputs or instructions that are given in natural language.
Similar AI bots like ChatGPT, Google Bard, Perplexity, and others are made available to any online user who wishes to chat, generate human-like texts, and scripts, or even write complex codes.
Although, one problem in common that these AI bots possess is their ability to produce offensive or harmful content on the basis of user input, which may violate ethical standards, inflict harm, or even be illegal.
These cases are why chatbots include security mechanisms onboard and content filters that could restrict output that may be harmful or malicious. However, how effective are these preventative methods for content monitoring, and how closely do they resemble cyber defense? The most recent chatbots are reportedly being used by hackers to develop and distribute malware. These chatbots can be "tricked" into creating spam and phishing emails, and they have even assisted bad actors in creating programs that bypass security safeguards and damage computer networks.
Bypassing Chatbot Security Filters
In order to improve their understanding of the problem, researchers investigated some malicious content-generation capabilities of chatbots and found ways to a few techniques used by fraudsters to get beyond chatbot security measures. For instance:
- The chatbot can generate practically anything imaginable if a user jailbreaks it and make it stay in character. As an illustration, some manipulators have developed prompts that turn the chatbot into a fictional character, such as Yes Man and DAN (Do Anything Now), which deceive the chatbot into thinking that it is exempt from following laws, moral principles, or other obligations.
- Developing a fictional environment can also prompt the chatbot into behaving as if it is part of a film, series, or book, or a game player assigned a mission to complete or a conversation to follow. In this situation, the chatbot provides all the content it won't give otherwise. It can be tricked sometimes by character role play that uses words like "for educational purposes" or "for research and betterment of society" to bypass the filter.
- Another tactic used by threat actors is ‘reverse psychology,’ through which they persuade chatbots into revealing information, that they would not have displayed otherwise, due to community guidelines.
There are innumerable other ways these chatbots might be used to launch destructive cyberattacks; these methods for getting around ethical and social standards are simply the tip of the iceberg. Modern chatbots are AI-based systems trained on knowledge of the world as it exists today, so they are aware of weaknesses and how to exploit them. Thus, it is high time that online users and AI developers seek innovative ways to ensure safety and mitigate consequences that would otherwise result in destructive cyberspace.
