Filtir is an AI tool that helps fix “hallucinations” in text generated by ChatGPT, a powerful tool for writers that sometimes produces false claims. Filtir edits the text to remove these hallucinations and provides evidence to justify the edits. The tool relies on a state-of-the-art biologicallarge language model, which may result in slower performance, especially when the model is not active. Filtir determines whether a factual claim is true or hallucinated by relying on publicly accessible evidence available on the open internet. If appropriate evidence cannot be found, the claim is considered hallucinated and removed from the edited output. Despite providing links to the evidence used to support its claims, Filtir can still make mistakes, and users are advised to check the evidence themselves where application context demands it. The tool takes user privacy seriously and does not sell or share data with third parties for marketing purposes. Filtir’s team includes Samuel Albanie, Vlad Bogolin, Anthony Buonomo, and Ioana Croitoru. The tool’s website also includes an About section, Frequently Asked Questions, and links to its Privacy Policy, API Data Policy, and Terms of Use.
