Ninth Circuit overturns injunction in Perplexity AI Amazon lawsuit over access
The Ninth Circuit Court of Appeals overturned a preliminary injunction against Perplexity AI, ruling that its agentic shopping assistant does not 'access' Amazon's servers under the Computer Fraud and Abuse Act. The court determined that the user, rather than the software tool, initiates the interaction, setting a significant legal precedent for how AI agents are classified under existing anti-hacking statutes.
Key Takeaways
- The court determined that Perplexity's Assistant is a software tool and not a 'person' capable of intentional access under anti-hacking statutes.
- Amazon failed to prove irreparable harm, with the court labeling its cyber-security risk arguments as weak.
- Technical evidence showed the Assistant sends browser screenshots to Perplexity servers rather than establishing direct server-to-server communication with Amazon.
- The ruling clarifies that user intent, not software automation, is the primary factor for attribution under the CFAA.
Why It Matters
This decision establishes a critical legal shield for developers of agentic AI tools by narrowly defining 'unauthorized access' in the context of automated browsing. By classifying the AI as a tool directed by the user, the court limits the ability of platform owners to use anti-hacking laws to block competitive shopping or data-gathering assistants. For the streaming and digital commerce ecosystem, this signals that existing statutes like the CFAA may be insufficient to prevent AI-driven scraping if the interaction is user-initiated. Industry observers should monitor the ongoing substantive proceedings to see if Amazon can establish liability through alternative contractual or misappropriation claims.
Additional Context
Perplexity AI has expanded its agentic shopping capabilities aggressively since the Ninth Circuit ruling, positioning itself as a direct competitor to traditional e-commerce interfaces. In early 2025, Perplexity launched its Comet browser with built-in AI agent features that can complete purchases on behalf of users, extending the same user-directed interaction model that the court found did not constitute unauthorized access under the CFAA. The company has also partnered with PayPal to enable direct checkout within its AI assistant, signaling that agentic commerce is moving from legal gray area toward mainstream infrastructure. These moves underscore why the Ninth Circuit's framing of AI agents as user tools rather than independent actors carries significant commercial weight beyond the immediate case.
The legal precedent intersects with broader regulatory scrutiny of AI-driven data collection and platform access. The U.S. Copyright Office published a report in January 2025 examining how AI systems interact with copyrighted content and recommending new frameworks for training data transparency, which could shape future claims against AI assistants that aggregate product information. Meanwhile, Amazon filed a separate lawsuit against Perplexity in November 2024 alleging that the AI company's web crawler ignored robots.txt directives, a claim distinct from the CFAA access question and still pending. That parallel proceeding tests whether contractual and technical access controls can succeed where criminal statutes did not, a distinction that will matter for any platform seeking to restrict AI agent behavior.
The ruling's technical reasoning also resonates with ongoing debates about browser automation and scraping in adjacent industries. In 2022, the Ninth Circuit's en banc decision in hiQ Labs v. LinkedIn held that scraping publicly available data likely does not violate the CFAA, and the Perplexity decision extends that logic to AI agents acting on user instruction. For streaming platforms that deploy recommendation engines and content discovery tools relying on third-party metadata, the precedent clarifies that user-initiated AI interactions with public-facing interfaces face a high bar for CFAA liability. However, legal scholars have noted that the ruling leaves open questions about terms-of-service enforcement and state-level computer crime statutes, which remain viable legal theories for platforms seeking to restrict automated access.
Read full article at inquisitiveminds.bristows.com
Enjoy our coverage?
Add StreamingMeme as a preferred source on Google to see more of our streaming news at the top of your Search results.
Add as preferred source