The concept of disinformation has long been contested, but the rise of generative artificial intelligence (Gen AI) introduces new challenges. In my talk, I would like to address them and present a new concept of disinformation that could deal with them. Firstly, disinformation is usually defined as misleading information intentionally created and disseminated. The rise of Gen AI undermines this, as LLMs do not mislead us intentionally (at least so far). They could generate disinformation if prompted to do so or even help users develop bots to disseminate it. However, it would be controversial to claim that they intentionally generate or disseminate misleading content. Additionally, certain researchers consider LLMs’ hallucinations as disinformation. Consequently, intentionality, as a characteristic of disinformation, is problematic in the era of Gen AI. Secondly, language cannot be used as a marker of disinformation, as AI-generated disinformation differs in its language patterns from human-created ones. Translators powered by Gen AI do not make the mistakes that previously helped us detect scams and phishing emails. Thirdly, due to Gen AI phenomena such as deepfakes, phishing emails, fraudulent reviews and testimonials, and scams, these came to the spotlight. Those can be covered by the term ‘disinformation’ but not by the term ‘fake news’, which is usually used just for misleading content in the form of news. Therefore, the term ‘disinformation’ is better suited to describe contemporary dangers than the term ‘fake news’. In conclusion, I argue that disinformation should be redefined primarily in terms of harm (actual or potential) and inauthentic dissemination, rather than intentionality.
Speaker: Zuzana Rybaříková