Recently, OpenAI has reached an agreement in a highly publicized copyright lawsuit, deciding to disclose the data used for training generative AI models to the plaintiff's attorneys.
Image source note: The image was generated by AI, and the image is authorized by the service provider Midjourney.
The plaintiffs in this lawsuit include several renowned authors such as Paul Tremblay, Sarah Silverman, Michael Chabon, David Henry Hwang, and Ta-Nehisi Coates. They filed a lawsuit against OpenAI and its affiliates last year, accusing the AI of using their works without authorization and generating text based on them, violating U.S. copyright law and state unfair competition law.
According to the ruling by U.S. District Judge Robert S. Irmas, the plaintiffs will gain access to a secure environment set up by OpenAI, where the inspection of training data is strictly limited. Recording devices are prohibited in the secure room, and OpenAI's legal team is also authorized to review any notes taken by the attorneys within. These measures make the disclosure of training data resemble a review of sensitive source code rather than a simple information sharing.
Although OpenAI insists legally that its use of copyrighted works falls under "fair use," this matter has attracted more attention because if OpenAI's training data is widely disclosed, it could lead to more legal actions. Currently, the copyright allegations against OpenAI not only come from these authors but also from other plaintiffs who are initiating similar lawsuits.
It is worth noting that in the future, more regulations may require AI developers to disclose their training data more transparently. The EU's Artificial Intelligence Act is expected to come into effect in 2025, requiring model providers to disclose detailed information about their training data to meet the legitimate needs of those concerned about their rights. Additionally, California has passed an AI data transparency bill, which has been signed by the governor.
Although OpenAI maintains that its generated content is based on an understanding of language, reasoning, and the world, there is still legal debate about whether the actions of AI models are appropriate. With an increasing number of lawsuits and legislative proposals emerging, the future of the AI field remains uncertain.
Key Points:
📝 OpenAI agrees to disclose training data to meet the needs of the copyright lawsuit.
🔒 Data inspection takes place in a strictly controlled secure environment, with recording devices prohibited.
⚖️ The future may face more regulations, promoting the demand for AI data transparency.