Perplexity responded to Amazon in the AI firm’s latest salvo against Big Tech.

Following in the footsteps of Reddit, the e-commerce giant issued a legal threat this week against Perplexity’s use of automated shopping agents.

Comet, which is Perplexity's AI-native browser offering, is able to automate online shopping upon user instructions, including product searches, price tracking, cart management, and even organize refunds and returns.

This didn’t go down well with Amazon, which argued in a blog “third-party applications that offer to make purchases on behalf of customers from other businesses should operate openly and respect service provider decisions whether or not to participate.”

In its cease-and-desist order, Amazon claimed Perplexity had refused to comply with multiple demands to “transparently identify Comet AI.”

The tech giant also claimed Comet “degrades the Amazon shopping experience” by not sharing critical product information, while also compromising customer data.

“Perplexity is intentionally evading Amazon’s identification of the Comet AI agent when it accesses the Amazon Store and thereby directly interfering with Amazon’s efforts to manage security risks," Amazon claimed. "This is particularly troubling given recent reports showing that Comet AI is vulnerable to prompt injection attacks, phishing, scams, and other forms of cyberattacks."

In response, Perplexity published a blog labeling Amazon’s actions as “bullying” that sets back AI innovation.

“This is Amazon’s first legal salvo against an AI company, and it is a threat to all internet users,” Perplexity wrote. “Amazon [is] more interested in serving you ads, sponsored results, and influencing your purchasing decisions with upsells and confusing offers.”

Perplexity also accused the firm of wanting “to eliminate user rights so that it can sell more ads right now and partner with AI agents designed to take advantage of users later.”

“It’s not just bullying, it’s bonkers,” Perplexity added, before repeating earlier assertions that AI agents are distinct from malicious crawlers, scrapers, and bots.

Agentic distinction and directories

The California-based business said the same in response to accusations from Cloudflare that it was obfuscating its crawlers to access content from the web giant.

Following the Cloudflare spat, Perplexity made another enemy in Reddit, landing a lawsuit accusing it of using various web intelligence platforms to unlawfully harvest data from its various forums.

Reddit accused Perplexity of bypassing its security measures across network traffic, leading to an example haul of close to three billion search engine results pages, including text, URLs, videos, and images.

In facing these various lawsuits, the onus is arguably on Perplexity to define how agents are dissimilar to scrapers and the like.

The rise of agentic AI meanwhile adds pressure on identifying agents within network traffic, a problem compounded by the lack of transparent verification for some agents as claimed in Amazon and Reddit's missives against Perplexity.

For their part, cybersecurity giants such as Palo Alto Networks and Cisco have made efforts to compile agent registries for users.

Another solution is the integration of model context protocol (MCP), the increasingly influential integration standard for AI agents, which has seen deployment by Zscaler, SUSE, and Cloudflare in recent months.