OpenAI, a pioneer in artificial intelligence, has strengthened its position around the globe with its state-of-the-art tech models. Renowned for its transformative interaction with individuals, its tools have been leveraged across several industries. Despite its enormous success, OpenAI’s influence has always sparked flummoxing debates over principles of ethics, intellectual property rights and regulatory frameworks. Recently, discussing the same principles, a New York Federal Court judge dismissed a copyright lawsuit brought against OpenAI by Raw Story Media Inc. and Alternet Media Inc.(Plaintiffs)
Pursuant to the Digital Millenium Copyright Act (‘the Act’) (DMCA) that implements WIPO’s two major treaties, the Plaintiffs brought this action against OpenAI over training data related issues. The Judge, Colleen McMohan of the Southern District Court of New York, dismissed the Plaintiff’s motion and denied them leave to re-plead as well. The Plaintiffs, who are news organizations, have alleged that ChatGPT by OpenAI does not possess any independent knowledge of the information provided in its responses and is trained using training sets by scraping the internet. They have also claimed that the Defendants have even scraped their copyrighted works of journalism without providing any Copyright Management Information (CMI), which is a violation of Section 1202(b)(i) of the Act. Thus, the Plaintiffs have prayed for actual or statutory damages along with an injunctive relief.
To render a decision, the Court delved deep into the nuances of the legislations revolving around the subject-matter. The Court looked at the following two legal standards:
i. Section 1202(b)(i) of the DMCA
This section, prohibits the removal or alteration of CMI without authorization. It emphasizes on a ‘double scienter’ legal requirement, meaning the Defendants must have actual knowledge of the removal of information and that it would reasonably lead to copyright infringement.
ii. Article III
This article requires the Plaintiff to demonstrate concrete, actual or imminent injury caused by the defendant and show that such an injury is only redressable by judicial relief. The Court also cited several cases to justify Article III, stating that Courts assess all reliefs sought by ensuring that such alleged harms have a close historical or common-law analogue that is traditionally recognized as actionable.
By citing the case of Trans Union, 594 U.S. at 435, the Plaintiffs have argued that if a risk of future harm is imminent and substantial, they are entitled to injunctive relief. The Defendants countered the same by stating that the Plaintiffs had failed to demonstrate any substantial risk of harm to their works, which the Court agreed. The Plaintiffs also provided statistics of previous instances of plagiarism by OpenAI, they failed to convince the Court that GPT’s output was likely to include a verbatim reproduction of their original content, owing to vast system data repository.
While dealing with Section 1202(b)(i) of the DMCA, the Court held that the major issue of the case was about use of journalism works without compensation, since OpenAI has entered into licensing agreements in the past with major copyright holders. The Court stated that such kind of an issue would not fall under the ambit of the said section and said under which statute such data use would be actionable, is an open question. The Court also noted that the Plaintiffs’ grievance did not constitute an elevated injury. Ultimately, the Judge dismissed the motion and allowed the Plaintiffs to file a new complaint, though she was sceptical about the fact that the Plaintiffs would not be able to allege a cognizable injury.
The dismissal of OpenAI’s copyright action in the United States raises serious concerns about AI and copyright law, particularly regarding the use of training data. In India, the legal framework governing AI-related copyright problems is currently emerging. Section 52(1)(za) of the Indian Copyright Act of 1957 permits “fair use” of copyrighted works for research and education, however AI training may dispute this clause, particularly when vast datasets are scraped without compensation or attribution. Similar to the US case, the question of whether AI platforms such as ChatGPT should compensate creators for their material remains an enigma. India has been at the centre of AI ethics issues, with previous incidents, such as the recent dispute of use of AI in examination by an LLM student of Jindal Global Law School., raising concerns about copyright violation and intellectual property rights. In all, the US judgement could heavily impact India’s approach to demarcate the scope of AI and copyright related infringements and to establish stringent regulations around the use of copyrighted content, especially as training data sets.
Authors: Seema Meena, Manasvi Shah & Simran Jayarao