One More Thing About Perplexity

Key Takeaways

• Perplexity seeks dismissal of direct infringement claims.
• Argues AI outputs are user‑prompted, not volitional.
• Creative story generation contradicts its “search‑only” narrative.
• Examples suggest extensive copyrighted training data usage.
• Case could reshape AI liability and fair‑use doctrine.

Summary

Perplexity AI has filed a motion to dismiss the New York Times and Chicago Tribune lawsuits, arguing that its answer engine merely follows user prompts and lacks the volitional conduct required for direct, contributory, or vicarious copyright infringement. The company characterizes its platform as a retrieval‑augmented search tool that synthesizes factual information from online sources, positioning the case as a fight over the future of open‑web search. However, user‑generated examples of original creative stories and stylistic rewrites reveal capabilities far beyond factual summarization, challenging Perplexity’s “passive conduit” defense. The dispute now pivots on whether these generative functions constitute actionable copying and how training‑data practices affect liability.

Pulse Analysis

The legal battle between Perplexity AI and major news publishers highlights a fundamental tension in how courts view generative artificial intelligence. While Perplexity frames its service as a modern search engine that retrieves and condenses factual data, the lawsuits focus on whether the platform’s outputs—especially those that replicate or transform copyrighted material—cross the line into infringement. By invoking the Cablevision precedent, Perplexity argues that the system is a passive tool activated solely by user prompts, a stance that could shield many AI services if upheld.

Recent user‑driven demonstrations, however, expose a different reality. When prompted to write an original story in the style of James Fenimore Cooper or to rewrite it as an Ian Fleming thriller, Perplexi­ty produced wholly new, stylistically nuanced prose. These creations are not mere summaries of existing web content; they are novel expressive works generated by the underlying large‑language model. Such outputs suggest a degree of creative agency that challenges the “no volitional conduct” argument and raises questions about the model’s training corpus, which likely includes extensive copyrighted literature.

The broader implications extend beyond Perplexity’s case. A court ruling that acknowledges generative AI’s capacity for original expression could expand liability for copyright infringement, compelling developers to disclose training data sources and implement stricter safeguards. Conversely, a decision favoring the passive‑tool defense might preserve the current permissive environment for AI innovation but leave creators of copyrighted works vulnerable. Stakeholders—from tech firms to publishers and policymakers—must watch this litigation closely, as its resolution will shape the future regulatory landscape for AI‑generated content.