You turned on Incognito Mode. Your chats went to Google and Meta anyway.
Incognito Mode is supposed to mean private. That's the whole point. You use it when you're searching for something you don't want tracked — health questions, financial decisions, personal research. The name itself is a promise.
A proposed class-action lawsuit filed in a San Francisco federal court alleges that Perplexity AI's Incognito Mode made that promise and didn't keep it. According to the complaint, conversations held in Incognito Mode were shared with Google and Meta — with personally identifiable information attached — in exactly the same way as regular conversations. The mode didn't protect anything. The lawsuit calls it a "sham."
This is a security alert for anyone who has used Perplexity to research anything sensitive.
What the lawsuit actually claims happened
The lawsuit was filed by an anonymous user identified as John Doe, a Utah man who used Perplexity as his primary AI search tool for serious personal matters — managing his family's taxes, getting legal guidance, and making investment decisions. He discovered, using browser developer tools, that his conversations were being transmitted to third parties.
The core allegation: Perplexity embedded tracking software — the kind of ad trackers that companies like Google and Meta use across the internet — directly into its search engine's code. The lawsuit describes these as "browser-based wiretap technology" that gives Google and Meta real-time access to what users type into Perplexity, including the full text of conversations.
The tracking happens at multiple points. Opening prompts — the very first thing you type — are always shared. Follow-up questions that appear as suggested clicks within the interface are shared when clicked. And for non-subscribed users, the initial prompt is shared with a URL that allows the entire conversation to be accessed by third parties.
The Incognito Mode failure is the most alarming specific claim. The lawsuit alleges that even paid subscribers who activated Incognito Mode still had their conversations transmitted to Google and Meta. Not anonymized data. Not aggregate statistics. The actual conversation content, along with email addresses and other identifiers that allow the recipient to personally identify the user.
Paid users paying for privacy protection got no privacy protection.
Why this is worse than typical ad tracking
Standard ad tracking is bad enough — websites tracking which pages you visit, building profiles of your interests, targeting you with relevant ads. Most people have come to grudgingly accept this as the price of free internet services.
What's alleged here is different in a meaningful way. People go to AI chat tools specifically because they want to ask things they wouldn't search for normally. Health symptoms they're worried about. Financial situations they're embarrassed to ask a human advisor about. Legal questions they can't afford a lawyer for. Relationship or family problems. Career anxieties.
The lawsuit explicitly mentions that users likely researched "sensitive topics" — health conditions, financial struggles, legal problems — with an expectation of privacy that the complaint alleges was never actually provided. When someone asks an AI about their debt situation, their symptoms, or their immigration status, that conversation carries a completely different privacy weight than a search for "best restaurants in Mumbai."
The lawsuit further alleges that Google and Meta can use this data for ad targeting and potentially resell it to additional third parties. Your conversation about tax troubles becomes a data point. Your health question becomes an audience segment.
Perplexity's response — and what it doesn't address
Perplexity's official response through a spokesperson was essentially a non-denial: the company said it hadn't been served any lawsuit matching the description and therefore couldn't verify the claims. This is a legal positioning statement, not a denial of the underlying allegations.
Meta's response pointed to its own policies stating that advertisers are not supposed to send sensitive information to Meta's systems. This is also not a denial — it's a shift of responsibility back to Perplexity as the party allegedly doing the sending.
Neither response addresses the specific technical mechanism alleged — that trackers are embedded in Perplexity's code and actively transmit conversations. Neither confirms or denies that the Incognito Mode functions as described in the complaint.
The India context — why this matters for Indian AI users
Perplexity has a significant and fast-growing user base in India. The AI search tool is popular among Indian professionals, students, and researchers precisely because it provides sourced, conversational answers to complex questions — tax queries, legal questions, investment research, medical information.
Indian users asking about GST filing, income tax obligations, property disputes, or health conditions through Perplexity are in exactly the category the lawsuit describes as most exposed. These are the kinds of sensitive, personal queries where the expectation of confidentiality is highest and where the consequences of that information reaching advertisers or data brokers are most significant.
There's also a financial angle specific to India. Many Indian users have been using AI tools to supplement or replace expensive professional consultations — asking Perplexity about legal rights, tax planning, medical symptoms, or investment strategies instead of paying a lawyer, CA, or doctor. If conversations about these topics were being transmitted to Google and Meta's ad systems, the breach of trust goes beyond just privacy inconvenience.
India's data protection framework is still evolving. The Digital Personal Data Protection Act 2023 is being implemented, but enforcement mechanisms for cross-border data flows involving AI tools remain unclear. Indian users who feel their data was mishandled have limited immediate legal recourse compared to US users who can join the class action.
What this means for AI search tools broadly
Perplexity isn't unique in using third-party trackers. The ad tracking ecosystem is deeply embedded in most web services. What makes this case notable is the specific combination of three factors: an AI tool that people use for sensitive personal queries, an explicit "Incognito Mode" privacy promise, and the allegation that the privacy promise was not honored even for paying subscribers.
The broader lesson is one that applies to every AI search and chat tool: the privacy guarantees these products advertise need to be independently verifiable, not just taken at the company's word. When a product is specifically used for sensitive personal questions — health, finance, legal, family — the privacy architecture needs to be transparent and auditable.
For users deciding which AI tools to trust with sensitive queries, the practical takeaway right now is: treat all AI search and chat tools as potentially logged and potentially shareable until proven otherwise. Use the most sensitive queries in tools with clearly documented, independently verified privacy policies. For truly sensitive matters — legal advice, medical situations, financial planning — the safest approach remains professional consultation with legally bound confidentiality protections.
TamilTech's take
The Perplexity lawsuit is a sharp reminder that "Incognito Mode" in any product means exactly what the company building that product decides it means — nothing more. There's no standardized definition. There's no regulatory requirement for what it must actually protect. When a company calls a feature "Incognito" and charges money for it while allegedly transmitting user conversations to Google and Meta with identifying information attached, that's not a privacy feature. That's branding. Indian users who have been using Perplexity for tax, legal, or medical research should take this seriously. Until these allegations are resolved — or refuted — treat sensitive AI queries as potentially not private.
Comments (0)
Be the first to comment!