Perplexity Updates: Comet, Publisher Deals and the Legal Docket

Amazon filed a first amended complaint on 21st Sep, after the Ninth Circuit vacated its injunction in August.

It keeps the two computer-access counts and adds a third, "Tortious Interference with Contractual Relations (California Law)".

The new allegation concerns the iPhone app. Amazon says that on Comet for iOS "Perplexity's own cloud servers take the user's Amazon authentication cookies" and "use those cookies to browse the Amazon Store itself".

That ran, Amazon alleges, "from Comet for iOS's launch on March 18, 2026, until at least May 11, 2026". It says the period overlapped with Perplexity's briefs to the appeals court that "no Perplexity computer ever has direct access to an Amazon computer".

Amazon also alleges the Comet agent carried the same user-agent string as Google Chrome. Its Agent Terms, quoted in the complaint, ask agents to include "Agent/[agent name]" in that string.

Perplexity's motion to dismiss of 13th Oct answers: "Two months ago, in this very case, the Ninth Circuit rejected the legal theory underlying Amazon's complaint."

It says "the desktop and iOS products work the same way in all legally relevant respects".

It asks the court to dismiss "with prejudice" and to stay discovery. The hearing is set for 20th Nov 2026 before Judge Chesney.

These are allegations and arguments, not findings. The same day, the judge referred Amazon's motion to compel discovery to a magistrate judge.

On 15th Sep, Judge Katherine Polk Failla signed a third amended civil case management plan. The lead case is 1:24-cv-07984 (S.D.N.Y.).

The order moves two deadlines. "All fact discovery shall be completed no later than December 10, 2026." Expert discovery, including depositions, runs to 16th Mar 2027.

Fact-witness depositions must finish by 10th Dec 2026. A telephone conference is set for 14th Dec 2026.

This is the third extension.

The original plan of 29th Aug 2025 ended fact discovery on 4th Jun 2026. The February amendment moved it to 20th Jul, and the May one to 15th Oct.

The trial estimate is unchanged. "This case is to be tried to a jury", the order says, with "the present best estimate of the length of trial" at 10 to 14 days.

Perplexity filed the request as a joint letter motion on 14th Sep, asking for more time to complete depositions. The order granting it says the court would enter the plan separately, which it did the same day.

Neither filing is readable through the free mirror. This entry rests on the docket text of the order itself.

No trial date is set. The case is now two years old with discovery still open.

There is no coverage of this anywhere. It is the biggest thing to happen in this docket in months.

On 21st Aug, Magistrate Judge Sarah L. Cave granted a sealing request from Encyclopaedia Britannica and Merriam-Webster. The case is Encyclopaedia Britannica, Inc. v. Perplexity AI, Inc. (1:25-cv-07546, S.D.N.Y.).

The plaintiffs had asked to file parts of a joint letter and four exhibits under seal.

The reason the order gives is the disclosure: the sealed passages contain "information revealed during on-site inspection of Perplexity's source code and highly confidential discussion of Plaintiffs' business strategy".

That is a discovery milestone rather than a ruling. The case has passed the stage where a plaintiff argues about what a system likely does.

Its lawyers and experts have now sat with the code. They did so under a Source Code Protocol and a Confidentiality and Protective Order the court entered earlier.

This claim turns on how a system retrieves, stores and reproduces text. For that, the source code is the evidence.

This order is the first confirmation on the record that Britannica's side has seen it.

Be clear about what is not knowable from here. The joint letter itself is a dispute the parties have put to the magistrate judge.

Its substance is redacted, and plaintiffs filed public versions with the confidential lines removed. Those files are not free to download from the docket.

We have read the order and not the letter. Nothing in the order says what the inspection found.

A sealing order is not a finding about anything.

The two motions to dismiss in this case remain briefed with no ruling. Our docket table has said that since it was built.

Discovery often runs while motions that could end a case sit undecided. That is normal rather than a signal.

Judge Milan D. Smith, Jr. wrote for the panel in Amazon v. Perplexity. He held Amazon unlikely to succeed on its Computer Fraud and Abuse Act and California CDAFA claims.

The reason: the USER, not Perplexity, "accesses" Amazon when Comet's assistant acts at that user's direction.

It is the first appeals-court ruling on whether an AI agent "accesses" a website. For now, it is a green light for agentic browsing.

The court's staff summary says the panel "vacated the preliminary injunction and remanded for further proceedings". Amazon's other claims survive on remand, so the case continues in the district court.

The district-court half of the fight had gone the other way. Judge Maxine Chesney granted Amazon a preliminary injunction on 9th Mar, five months before the appeal landed.

Why it matters beyond these parties: must an AI agent identify itself, or does it carry its user's permissions? That is the open question of the agentic web.

The EFF, which backed Perplexity's position, reads the ruling as protecting browser-building generally. Publishers argued the opposite as amici: Digital Content Next told the court that "advertisers pay for their ads to be seen by humans, not AI agents".

Judge Paul A. Engelmayer largely denied the motions to dismiss in Reddit v. SerpApi, Perplexity and others. His opinion runs 63 pages.

The DMCA section 1201(a) circumvention claims proceed against both Perplexity and SerpApi.

Two holdings carry the weight. Google's SearchGuard was held to be a measure that controls access to copyrighted work.

Reddit was held to have standing. The court wrote, as quoted by MLex, that Reddit "epitomizes the 'global digital on-line marketplace for copyrighted works'" the DMCA protects.

Per Press Gazette's tracker, the judge found Reddit "has plausibly pleaded that Perplexity conspired with at least one of the three data scrapers to bypass access controls and obtain Reddit's material".

Not everything survived. A trafficking claim against SerpApi was dismissed, along with the state-law unfair-competition and unjust-enrichment claims.

Oxylabs' separate motion is still in briefing through late August.

Reddit's core theory comes from its October 2025 complaint. It says the defendants scraped Reddit content out of Google's search results rather than from Reddit itself.

Reddit calls that data laundering.