Free AI Isn't the Enemy. Putting the Wrong Things in It Is.
Perplexity raised $1 billion on the privacy pitch. The lawsuit filed yesterday tells a different story.
Free AI Isn’t the Enemy. Putting the Wrong Things in It Is.
Hey there,
I want to tell you about a guy in Utah.
He used Perplexity AI the way a lot of service business owners use it: for research, for thinking through problems, for strategy work he didn’t want anyone else seeing. He typed in details about his family’s finances. His tax obligations. His investment portfolio. His personal financial strategy.
He assumed the conversation was private.
The lawsuit filed Tuesday in federal court in San Francisco says it wasn’t.
According to the complaint, Perplexity embeds tracking scripts that activate the moment you log in and transmit complete transcripts of your conversations to Meta and Google in real time. Not summaries. Not anonymized data. The full text of what you typed, going straight to the two largest advertising platforms on the internet.
The case is Doe v. Perplexity AI Inc., 3:26-cv-02803. The plaintiff is listed as John Doe. The irony is not subtle.
And here’s the kicker. Perplexity has an Incognito mode. They built it specifically to signal that they were different from surveillance-based search engines. The lawsuit alleges the trackers work right through it.
Perplexity raised over $1 billion on the positioning that it was the privacy alternative to Google. When asked about the lawsuit this week, their spokesperson said they “haven’t been served” and therefore “cannot verify its existence.” They did not deny the tracking.
Meta’s response was to point reporters to their policy page saying advertisers shouldn’t send them sensitive data. One outlet described this as a casino pointing to its responsible gambling brochure.
Google said nothing.
So why does this matter to you specifically.
The man in Utah typed financial strategy into a free AI tool. Think about what you typed into a free AI tool this week.
Client names. A pricing proposal you were workshopping. Competitive intel you wouldn’t share with your team. A vendor negotiation you were rehearsing. Your own revenue projections. A problem you’re having with a client that you worked through out loud with an AI chatbot because it felt safer than telling anyone who knows you.
Every one of those conversations exists somewhere. And if the platform running the tool has a business model built on data, which every free platform does, then “somewhere” is doing a lot of work in that sentence.
This is not a Perplexity-specific problem, even though the lawsuit is. And I want to be direct about something: free AI has been genuinely game-changing for small business owners. People who couldn’t touch this stuff two years ago are now running automations, producing content, and serving clients at a level that used to require a full team. Use it. It’s incredible.
Just know the trade. Free products run on something, and for most of them, that something is your data. You already know this from years of using Google and Facebook for free. The same deal applies here, and the only difference is that what you’re putting into AI tools is usually a lot more sensitive than your Instagram habits.
Here is what I actually think you should do, right now, before you open another AI tab.
Know which tier you’re on and what it means. A $20/month subscription changes the business model. It doesn’t guarantee privacy, but it changes the company’s incentives significantly. If you’re doing sensitive client work, the paid tier of the same tool you already use is often the right answer, not a different tool entirely.
Treat free AI tools like public Wi-Fi. Use them for things you wouldn’t mind being visible. Never type client names, financial details, pricing strategy, or anything you wouldn’t want a competitor or a journalist to see. The research tab is fine. The strategy session is not.
“Incognito mode” on an AI product means almost nothing. Browser incognito hides your history from other people using your computer. It doesn’t hide you from the website. AI incognito mode in many cases means even less. Assume the platform sees what you type, full stop.
Check your AI tool’s data training policy. Most major platforms let paid enterprise users opt out of having their conversations used to train future models. Free tier users often cannot opt out. If you’re doing client work on a free tier, that conversation may be training the next version of the model. For a lot of service businesses, that is not a theoretical risk.
The lawsuit might go nowhere. That’s genuinely possible. But that’s not why I’m telling you this. I’m telling you because the behavior in the complaint is what happens when you hand sensitive information to a free product without thinking about what you actually agreed to. You’ve already figured this out with Instagram and Facebook. Your AI tools deserve the same skepticism, and the information you put in them is usually way more valuable than anything on your social feed.
A big part of what I help business owners sort through is exactly this: not just which AI tools to use, but how to use them without handing your client relationships and business strategy to platforms that have zero obligation to protect it.
If you want to talk through what your current AI stack looks like from a privacy and cost standpoint, I’m opening a small number of spots for strategy sessions in April. Or just hit reply and tell me what tools you’re actually using for client work. I read every one.
Talk soon,
Andrew