TL;DR:
The Ninth Circuit on August 4, 2026 vacated a district court’s preliminary injunction blocking Perplexity AI’s agentic web browser from being used on Amazon’s site. The court held that Amazon was unlikely to prove CFAA liability because Perplexity did not “access” Amazon computers; instead the user, aided by Perplexity’s Assistant, drove the access. The decision remanded for further proceedings. For trial teams, this narrows the likelihood of CFAA injunctive relief when AI tools act through user-directed agents rather than as direct system access by the defendant. Practitioners should recalibrate strategies for market-blocks, data access disputes, and AI-driven discovery, and consider alternative theories when seeking urgent relief. Objection Academy can help attorneys prepare for AI-influenced evidentiary disputes and objections in these evolving contexts.
Background and holding
Amazon.com Services, LLC v. Perplexity AI, Inc. concerns whether Perplexity’s AI-enabled web browser, Comet, and its AI “Assistant” can trigger CFAA liability when used to access content on Amazon.com. The Ninth Circuit’s August 4, 2026 opinion concludes that the district court erred in granting a preliminary injunction because Amazon is unlikely to prove that Perplexity itself accessed Amazon computers for CFAA purposes. The court emphasizes that it was the user, not Perplexity, who accessed Amazon using the Assistant, so the defendant did not itself violate the CFAA as a direct actor. The panel vacated the district court’s injunction and remanded for further proceedings. This summary, based on the court’s published opinion, reflects the court’s articulation of the CFAA’s “access” prong and the equitable balance applied to injunctions. (cdn.ca9.uscourts.gov)
Crucially, the Ninth Circuit’s analysis distinguishes between an actor who directly “accesses” a computer and a tool that enables a user to perform access. The court’s decision forecloses a straightforward theory that AI browsing bots, acting as agents, automatically expose the tool provider to CFAA liability simply by enabling the user’s actions. Instead, liability turns on who is the actual accessor under the statute and whether the defendant exceeded authorized access or intentionally accessed a protected computer. The opinion was issued on August 4, 2026, and the court’s docket confirms the decision’s publication and timing. (cdn.ca9.uscourts.gov)
What the decision means for trial teams
- Narrowed path to CFAA injunctions in AI contexts: The decision signals that in many AI-enabled access disputes, courts may require the entity to be the actual accessor or to have directly caused access to a protected computer. If the AI tool functions as a mediator for the user, the defendant’s liability under CFAA and CDAFA may be weaker at the injunction stage. This reduces the likelihood of securing rapid, tool-blocking injunctions in AI browsing disputes absent another theory. (cdn.ca9.uscourts.gov)
- Emphasis on who controls access: Litigants pursuing CFAA relief should carefully map who initiates the access, who operates the accessing mechanism, and whether the defendant actually accessed the target system or merely supplied a mechanism that a user employed. The court’s reasoning centers on the distinction between user action and tool action. Practitioners should gather and preserve evidence about the user’s direct interaction with the target site and the role of the AI assistant in that interaction. (cdn.ca9.uscourts.gov)
- Remand means ongoing proceedings: With the injunction vacated and the case remanded, trial teams should anticipate additional proceedings that may revisit evidentiary issues, the scope of CFAA claims, and remedies. The procedural posture suggests that posture and record development on remand will shape whether any injunction can be tailored to address a particular misuse without broad, blanket prohibitions. (cdn.ca9.uscourts.gov)
Practical implications for trial strategy
- Reframe urgent relief when AI tools are involved: If a client seeks to stop or restrict an AI tool’s access to a platform or data source, the strategy may need to pivot toward non-CFAA avenues or to focus on direct access by the user. Consider contract-based or terms-of-use violations, or other statutory theories that might support injunctive relief where CFAA arguments face headwinds after this decision. (cdn.ca9.uscourts.gov)
- Evidence collection and witness preparation: Given the decision’s emphasis on who accessed the system, trial teams should emphasize preservation and authentication of AI-assisted access events, including logs from user interactions with AI tools and any server-side records that show user-initiated access. In later stages, this can influence both liability theories and the framing of equitable relief. (cdn.ca9.uscourts.gov)
- Trial-readiness for AI-driven disputes: As AI-enabled discovery, evidence generation, and browsing become more common, litigants should prepare for objections and admissibility challenges tied to machine-generated content. While the Ninth Circuit case does not establish a rule about AI-generated evidence itself, it underscores the evolving battlefield where AI acts as an instrument of access. Objection Academy offers structured practice to handle AI-related objections and evidentiary challenges in real courtroom settings, helping trial teams stay crisp on objections to AI-sourced material when it matters most. (uscourts.gov)
Strategic actions for practitioners now
- Audit CFAA and CDAFA theories of liability: Reassess whether injunctive relief can be obtained by targeting the actor who actually accessed the site and whether alternate theories better fit the case posture. Review district court orders and appellate opinions for how access and authorization are argued in AI contexts. The Ninth Circuit opinion provides a clear framework to scrutinize “who accessed” before seeking relief. (cdn.ca9.uscourts.gov)
- Prepare for remand development: Anticipate a more focused course on the access theory during remand proceedings, with updated evidentiary submissions and potential narrowing of relief while preserving the ability to address legitimate site-protection interests. (cdn.ca9.uscourts.gov)
- Leverage training resources for AI-era advocacy: In parallel with case developments, trial teams should sharpen objections to AI-derived or AI-assisted evidence and develop seamless trial-readiness workflows. Tools and training from Objection Academy can support trial teams in shaping persuasive objections to AI-generated content and in maintaining rigorous evidentiary standards in fast-evolving digital contexts. (uscourts.gov)
Where to monitor next
Stay attuned to any further developments on remand in Amazon v. Perplexity AI, including how the district court addresses equitable factors and any new evidence or arguments introduced on remand. The Ninth Circuit’s August 4, 2026 decision stands as a clarifying moment in CFAA injunction practice for AI-enabled access disputes and will likely influence subsequent trials where AI agents operate as intermediaries in online access.
Sources:
- Amazon.com Services, LLC v. Perplexity AI, Inc., Ninth Circuit, Aug. 4, 2026, No. 26-1444. Opinion available as published PDF from the Ninth Circuit. turn3view0
- Ninth Circuit Opinions — Amazon.com Services, LLC v. Perplexity AI, Inc. (CA9 2026) list of decisions, August 2026. turn2search1
- Court coverage: CourtListener/Courthouse News summaries and Law360 reporting on the August 4, 2026 ruling. turn2search12, turn2search8
- Federal rules and AI-evidence rule context (for related practice considerations): Advisory Committee materials and related coverage. turn1search12, turn0search2
- Objection Academy (for AI-related evidentiary objections and trial readiness references). turn0search9