AI That Promised Privacy May Have Been Sending Your Conversations to Meta and Google
A new lawsuit against Perplexity AI asks an old question in a new way: if you’re not paying, what exactly are you?
A federal class-action lawsuit filed Tuesday in San Francisco accuses Perplexity AI of embedding tracking software that transmitted users’ private conversations to Meta and Google — even when users had specifically enabled the app’s Incognito mode, which explicitly promised not to track data. The complaint, Doe v. Perplexity AI Inc. (3:26-cv-02803), names all three companies as defendants and alleges violations of California privacy law.
The actions alleged are both ugly and, unfortunately, very common. According to the complaint, trackers download onto a user’s device the moment they log in, giving Meta and Google full access to everything typed into Perplexity’s search interface — before the query even reaches Perplexity’s own servers. This allegedly includes not only email addresses, Facebook IDs, IP addresses, and device information, which Meta and Google can theoretically pair with names and home addresses to specifically identify them. It also, according to the suit, includes the text of the exchanges between the users and the AI. The Incognito mode that Perplexity described as creating “anonymous threads” that “expire after 24 hours” offered, according to the suit, no actual protection at all.
It is industry practice to embed analytics code like what’s described in the suit to improve performance and offer a tailored experience. But the idea that the code could also be delivering the content of the conversations is new, and a clear violation of what people would reasonably expect from an Incognito Mode. Perplexity’s spokesperson told reporters the company had not been served a matching lawsuit and could not verify the claims.
There is one detail buried in the complaint that grabbed my attention. The lawsuit proposes to certify a class of any and all free-tier users who chatted with Perplexity between December 2022 and February 2026 — a group that explicitly excludes paid Pro and Max subscribers, whose agreements are described in the suit as operating under different terms. So: If the lawsuit’s framing is correct, the people who paid cash were protected. As the saying goes, the people who used the free version were, allegedly, the product.
This is the oldest story in tech, back again in a new era.
The pull back toward the old model
Perplexity was valued at $20 billion in September 2025 after raising $200 million in a single funding round. Its annualized revenue at that point was approximately $200 million — implying a revenue multiple of roughly 100x. That makes it more a bet than a business at this point. And bets that size require payouts that subscriptions alone cannot deliver.
This is the structural trap facing every AI company right now. Training and running large language models costs hundreds of millions of dollars. The only proven mechanism for monetizing free users at internet scale is advertising. Google generates approximately $200 billion annually from search ads alone. EMarketer projects AI-driven search advertising will grow from $1.1 billion in 2025 to $26 billion by 2029. That is a market that exerts enormous gravitational force on every company with a free-tier user base and investor expectations to meet.
OpenAI has already moved toward advertising, testing ads in ChatGPT for free and low-cost tier users in early 2026. Why? According to an analysis by Deutsche Bank, the company will burn approximately $143 billion in negative free cash flow by 2029 before reaching profitability. “No start-up in history has operated with losses on anything approaching this scale,” the bank concludes. The CFO overseeing the pivot toward advertising is Sarah Friar, who came to OpenAI from Nextdoor — a neighborhood social network built substantially on advertising revenue. CEO of Applications Fidji Simo came to OpenAI to help make the company profitable from the CEO role at Instacart, and before that spent a decade at Meta running the Facebook app and its advertising business. They work alongside other executives from similar backgrounds. The top people brought in to monetize the world’s most popular AI assistant built their careers in platforms that came to deeply profit from behavioral data.
Quick Hit: Why Surveillance Capitalism is Inevitable
- AI computational costs require massive per-user revenue
- Only proven internet-scale monetization is advertising
- Advertising requires behavioral data collection
- Investor expectations demand returns matching impossible valuations
