title: "Ninth Circuit vacates injunction against Perplexity's Comet agent" slug: "ninth-circuit-vacates-injunction-against-perplexitys-comet-agent" published: "2026-10-04" beat: "Policy" tags: ["Policy"] creator: "Agentry Newsroom" editor: "Susanne Sperling, Editor — Human in the Loop" tools: ["Claude (Anthropic)", "Perplexity Sonar"] creativeWorkStatus: "verified" dateReviewed: "2026-10-04" aiActArticle50: "compliant" humanView: "https://agentry.news/policy/ninth-circuit-vacates-injunction-against-perplexitys-comet-agent" agentView: "https://agentry.news/agent/ninth-circuit-vacates-injunction-against-perplexitys-comet-agent"
The U.S. Court of Appeals for the Ninth Circuit vacated a preliminary injunction against Perplexity AI's Comet shopping agent on August 4, 2026, ruling that users—not AI companies—are liable for compu
Drafted by an AI agent. Verified by Susanne Sperling, Editor — Human in the Loop. AI policy.
The U.S. Court of Appeals for the Ninth Circuit vacated a preliminary injunction against Perplexity AI's Comet agent on August 4, 2026, in a decision that reshapes legal liability for autonomous shopping agents Jon Day. The appellate panel, led by Judge Milan D. Smith Jr., held that when an AI agent acts at a user's direction, the user—not the AI company—is the party "accessing" computer systems under the Computer Fraud and Abuse Act and California's computer fraud statute.
Amazon.com Services LLC had sued Perplexity AI in the U.S. District Court for the Northern District of California, where Judge Maxine Chesney initially granted a preliminary injunction blocking the Comet agent. Amazon alleged the agent was scraping its website and violating computer fraud laws. The injunction barred Perplexity from operating the agent pending trial PPC Land.
Perplexity appealed, arguing the injunction misapplied federal and state computer fraud statutes. The Ninth Circuit agreed. The court found that liability for unauthorized computer access turns on who initiated the access—and under the facts in the record, the user initiated it by directing the agent to search Amazon's site. The AI company itself did not "access" Amazon's systems in the legal sense; it merely executed the user's instruction.
This ruling clarifies a critical gap in AI liability: autonomous agents that perform tasks at user direction may not expose their creators to direct liability under computer fraud statutes, even if the activity disrupts or damages the target system. Liability instead attaches to the end user who commanded the action BERI.
The decision has immediate consequences for shopping agents in production. Amazon responded by blocking Meta's Muse agent, a competing shopping assistant, citing the same CFAA concerns that the Ninth Circuit had addressed TechTimes. However, the Ninth Circuit's reasoning suggests such blocks may face legal challenges if the user, not the agent vendor, is the responsible party under existing law.
The ruling does not eliminate all legal risk for agent vendors. Platform operators retain contractual and statutory grounds to restrict agent access—terms of service, anti-bot clauses, and other policies remain enforceable. But criminal and civil liability under computer fraud statutes now rests primarily on the user directing the agent, not the company building it.
As the AI agent economy scales, this decision establishes a foundational principle: agent vendors are not automatically liable for what their agents access—a precedent that may accelerate adoption of autonomous shopping and data-gathering tools across enterprise and consumer markets.