Amazon.com Services, LLC v. Perplexity AI, Inc.

Amicus Brief

Issue: In *Amazon.com Services, LLC v. Perplexity AI, Inc.*, the ACLU, ACLU of Northern California, and Knight First Amendment Institute argue that the Computer Fraud and Abuse Act does not reach an AI-powered browser that accesses platform data on behalf of authenticated, consenting users. The brief presses the non-obvious question of whether a platform's unilateral cease-and-desist letter can convert user-delegated access into criminal unauthorized access — and whether any CFAA construction that permits platforms to define their own liability triggers by sending demand letters would unconstitutionally chill automated journalism and public-interest research.

The ACLU, ACLU of Northern California, and the Knight First Amendment Institute at Columbia University filed this amicus curiae brief at the Ninth Circuit (No. 26-1444) with consent of all parties under FRAP Rule 29, in support of Defendant-Appellant Perplexity AI and in favor of reversing the district court's preliminary injunction. The brief argues that when an AI tool acts at a logged-in user's direction using that user's credentials, it operates as an extension of the user's own authorized access, rendering a platform's separate objection to the AI entity legally irrelevant under the CFAA. Relying on *Van Buren v. United States*, *hiQ Labs v. LinkedIn*, and *Facebook v. Power Ventures*, amici contend that CFAA authorization turns on technical access gates — not on a platform's unilateral written demands — and that authenticated access to publicly viewable data does not become unauthorized merely because the platform objects to who is retrieving it. The brief also advances a constitutional avoidance argument, documenting that an expansive CFAA reading would chill adversarial platform research and automated investigative journalism, invoking *Sandvig v. Barr* and citing concrete examples including the NYU Ad Observatory and litigation against the Center for Countering Digital Hate.

This brief pushes the Ninth Circuit toward a significant doctrinal extension of *hiQ Labs* — moving that decision's public-data logic into the contested terrain of authenticated, user-delegated AI agent access, a question no circuit has cleanly resolved. If the court accepts the user-authorization-as-delegation framework, it would effectively insulate a broad class of AI browsing and research tools from CFAA liability so long as they operate with a user's credentials and consent. The brief's treatment of *Facebook v. Power Ventures* is the argument's most vulnerable point: that decision specifically permitted CFAA liability to attach after an individualized cease-and-desist, and Amazon's stronger theory — that Perplexity was never independently authorized in the first place — maps more naturally onto *Power Ventures* than amici acknowledge. The constitutional avoidance thread is nonetheless significant: even if the textual argument fails, a ruling that endorses the chilling-effect analysis could constrain how broadly any CFAA holding is written. The case is worth watching as an early test of how appellate courts will apply *Van Buren*'s gates-up/down framework to AI agents acting on behalf of human users.

Issue: Insufficient text to determine — the summons identifies Amazon.com Services LLC as plaintiff and Perplexity AI, Inc. as defendant but does not disclose the specific legal claims, statutes, or theories of liability asserted in the underlying complaint.

On November 4–5, 2025, Amazon.com Services LLC filed a civil action against Perplexity AI, Inc. in the Northern District of California (Case No. 3:25-cv-09514-SK), and a summons was issued requiring Perplexity to answer or move under Rule 12 within 21 days of service. The document before the court is the summons and its proof-of-service form; no complaint, motion, or substantive order is included. The case was subsequently referenced in a filing dated November 11, 2025 under a related docket entry (Document 16). No rulings, arguments, or standards have been set forth in this document.

Insufficient text to determine — the summons alone reveals only the identity of the parties and the forum, not the legal theories that would bear on platform liability, First Amendment doctrine, or AI regulation.