April 1, 2025:
OpenAI Faces Allegations of Misusing OReilly Content - A report by the AI Disclosures Project accuses OpenAI of using paywalled OReilly books without permission to train its GPT-4o model, which shows greater recognition of these texts than previous models. The report used DE-COP, an inference attack method, to suggest GPT-4o was likely trained on copyrighted material.
While the findings aren't conclusive, they raise ethical concerns about OpenAI's data practices. OpenAI, which did not comment, is already facing several lawsuits over its handling of copyrighted data.
