The San Francisco-based 9th U.S. Circuit Court of Appeals has dealt a significant blow to Amazon's legal strategy by overturning a temporary ban that had prevented Perplexity from deploying its AI-powered shopping agents on the e-commerce giant's platform. In a decision handed down on Tuesday, the appeals court found that Amazon was unlikely to succeed in proving that Perplexity's technology violated federal computer fraud and abuse statutes, marking the first major appellate ruling to address whether artificially intelligent agents can legally access and interact with commercial platforms on behalf of users.
The case represents a watershed moment for the emerging field of agentic AI—sophisticated systems capable of planning, reasoning, and executing tasks with minimal human intervention. These tools have become increasingly sophisticated and commercially valuable, with companies developing agents that can browse the internet, make purchases, and conduct transactions autonomously. The legal framework governing how these systems interact with existing online platforms has remained murky, making the appellate court's decision particularly significant for the entire technology sector and for companies developing or considering deployment of such technologies.
Amazon initiated legal proceedings against Perplexity in November, alleging that the startup had engaged in unauthorized access to customer accounts through its Comet browser and associated AI agent functionality. According to Amazon's complaint, Perplexity's system could log into users' shopping accounts and execute purchase orders without proper authorization, creating what the company characterized as serious security vulnerabilities. Amazon further asserted that it had repeatedly requested Perplexity cease this activity, and that the startup had disregarded these warnings.
In March, a federal judge in California granted Amazon a preliminary injunction that temporarily halted Perplexity's use of agentic AI on Amazon's platform. The judge determined that Amazon had presented substantial evidence suggesting Perplexity's operations violated the Computer Fraud and Abuse Act, a sweeping 1986 statute that criminalizes unauthorized computer access. This initial victory appeared to establish a strong position for Amazon in what many expected would be a lengthy legal battle.
Perplexity's defense strategy centered on a technical but crucial distinction in how the law defines "access." The startup argued that its users—rather than Perplexity itself—were the ones accessing Amazon's platform through the AI agents. Under this interpretation, the users possessed authorization to access their own accounts, and therefore Perplexity could not be held liable for unauthorized access. The company additionally framed the dispute as an attempt by Amazon to prevent users from choosing their preferred AI tools, characterizing the lawsuit as anticompetitive.
In reversing the preliminary injunction, the appellate court embraced Perplexity's legal reasoning. The court concluded that under the Computer Fraud and Abuse Act, it is the users deploying the AI agents who technically "access" Amazon's platform, not Perplexity as an independent actor. This distinction proved decisive, as the court found that since the users possessed legitimate authorization to access their own accounts, Perplexity could not be prosecuted under the statute. The ruling essentially established that AI intermediaries acting on authorized users' behalf may not themselves be liable for computer fraud simply by facilitating that access.
Amazon responded to the defeat with a statement indicating the company would not accept the decision passively. An Amazon spokesperson declared the company remained "confident in our case" and signaled that additional legal strategies were under consideration, though the company did not specify what form those next steps might take. The setback appears to have rattled Amazon's confidence in the legal theories it had relied upon to that point.
Perplexity celebrated the decision as vindication of its position. Company spokesperson Jesse Dwyer emphasized the startup's commitment to defending users' rights to select their preferred AI tools and expressed confidence that the broader legal proceedings would ultimately prove favorable to the company. Dwyer's statement contained a subtle jab at Amazon, suggesting that the e-commerce platform had sought to block users from accessing AI agents that bypass Amazon's advertising infrastructure—implying that the lawsuit was motivated by commercial rather than security concerns.
The ruling carries substantial implications for how agentic AI systems can operate within the existing internet ecosystem. If the appellate court's logic holds firm in future cases, it establishes that companies developing AI agents need not fear liability under computer fraud statutes so long as the underlying users possess authorization to access the platforms being targeted. This could accelerate development and deployment of increasingly autonomous AI systems designed to interact with websites, retail platforms, and other digital services.
For Malaysian technology companies and investors tracking developments in artificial intelligence, this decision suggests that regulatory barriers to agentic AI deployment in major markets may be lower than previously anticipated. The ruling could influence how platforms in Southeast Asia approach similar AI technologies, particularly as regional companies develop their own agentic AI capabilities. However, the decision does not preclude other legal theories—such as trademark infringement, terms of service violations, or data protection claims—that companies like Amazon might pursue in subsequent litigation.
The broader tech industry is watching closely as regulators and courts grapple with how existing legal frameworks apply to AI systems that operate with increasing autonomy. This San Francisco appellate decision represents a clear statement that courts may be inclined to interpret traditional computer fraud statutes in ways that accommodate AI agents acting on users' behalf, provided those users maintain legitimate access rights. Whether this approach ultimately survives further legal challenges remains to be seen, but the ruling has temporarily cleared a significant legal hurdle for companies developing and deploying agentic AI technologies at scale.
