OpenAI Inc. is placing the concept of “prompt hacking” at the forefront of its defense against an array of copyright lawsuits filed by publishers and authors. The company contends that the plaintiffs have utilized deceptive and extensive manipulation of ChatGPT to produce outputs that allegedly infringe on copyrighted works. In an effort to substantiate these claims, OpenAI, often named as a co-defendant with Microsoft Corp., has requested broad discovery in several cases to uncover the exact nature of the prompts that led to the contentious outputs.
This discovery strategy aims to reveal plaintiffs’ extensive efforts to manipulate the AI tool, portraying the putative infringing outputs as “highly anomalous results” that stem from tens of thousands of attempts and sophisticated prompts. This approach is intended to persuade juries and legal practitioners that the outputs in question are not indicative of ChatGPT’s general functionality, but are instead the result of deliberate attempts to elicit specific, protected content.
Access to the plaintiffs’ chatbot prompts is sought as a crucial element to bolster OpenAI’s defense in these copyright suits, emphasizing that the disputed outputs could only be generated through intentional manipulative tactics. For legal professionals monitoring this development, the unfolding dynamics have significant implications for how AI-generated content is evaluated in copyright disputes.
For further details, read the full report on Bloomberg Law here.