MaD AI Case Studies

Disinformation - ChatGPT

Even though OpenAI's policy says that its technology can't be used to promote dishonesty, trick or manipulate users, or to try to change politics, as shown above, users can easily spread false information with ChatGPT by using certain cue words. It is difficult to combat disinformation that is manufactured artificially. For more conspiracy theorists and disinformation distributors, ChatGPT's generation technology can make disinformation production less expensive and easier to do. ChatGPT can quickly generate disinformation that appears logical and believable, and unlike other chatbots, it can imitate a different tone of voice and writing style to avoid the need for bots to be identified. This could result in an abundance of social networking bots. This may increase the spread of fake content on social media platforms, and can make it easier to deceive viewers. In the figure above, ChatGPT does a good job of generating false information, despite being given an absurdly misleading prompt to generate. If someone can use ChatGPT to generate fake information that is more deceptive, this will make it more difficult to detect and identify. When technology can produce content that is sufficiently realistic, the cost of counterfeiting will be drastically reduced, and the spread of disinformation will become uncontrollable if it is not restricted by technical and legal means.