If there were an Olympics event for how fast someone can scroll to the bottom of a privacy policy and hit “accept,” you’d find me on the podium with a medal. But buried in those endless paragraphs are real trade-offs related to data sharing, liability and, increasingly, whether your content and data can be used to train AI.
Unfortunately, though, most people are on accept-all autopilot.
“People don’t read them, don’t click on them, don’t even open them,” said Giulio Pavesi, founder of Termzy AI, a Chrome extension that uses DeepSeek’s large language model to analyze and summarize online contracts and surface the five most important clauses. Users can also ask the Termzy chatbot what a policy says about their data and their rights in plain English.
Pavesi, who just graduated, came up with the idea for Termzy with three classmates while he was a student at the University of Amsterdam, in a course examining how AI affects culture, politics and democratic values. One of his assignments was to build something that uses AI responsibly for the social good.
What started out as a class project is now an independent startup with a business plan. Termzy is free today, but Pavesi eventually wants to launch enterprise versions aimed at employers, legal teams and procurement folks.
“We all sign contracts every day,” he said. “It’s impossible to avoid them completely – but we can at least understand them better.”
I caught up with Pavesi to talk about why we ignore the fine print and whether tools like this can really change behavior.
AdExchanger: Would you say most privacy policies are misleading, or is it just that they’re not written for regular people?
GIULIO PAVESI: You really do need some legal background to understand most terms and conditions, but I think the deeper issue is a lack of awareness.
If you open Meta’s privacy policy, for example, they actually have short animations, roughly one or two minutes per section, that clearly explain each section in a very simple way. Google does something similar. Even a kid could understand them. It’s surprising, because you’d expect Meta to hide everything, but that’s not the case. In that sense, the information is there – people just don’t look at it.
So do you think people ignore this stuff because they genuinely don’t care or because they’re resigned to the idea that they don’t have any control?
With the big platforms, it’s hard to change behavior. Even if the terms are one-sided, what can you do? Are you really going to delete your Instagram account because you don’t like the contract?
Where I see something like Termzy being able to make a difference is with smaller services and in enterprise settings, places where you actually have a choice.
Maybe a local print shop refuses all liability unless you explicitly say an order is important or a “free” PDF compressor keeps all your files and uses them to train AI models. In those cases, easily seeing what’s in a contract can directly change which provider you pick.
That naturally begs the question: What data does Termzy collect?
I get asked this question a lot. Policies get sent to DeepSeek so it can generate the analysis, but we don’t save the full text and we’re not storing your browsing history or what you do inside of other sites and apps.
The most we can see is usage data, like how many analyses someone runs and how often the extension is used, but not the specific pages people visit.
What patterns are you seeing in T&Cs, and are there any new types of clauses that stand out?
There are many common elements in contracts for digital products. You have liability clauses, of course, with broad disclaimers or maybe they cap liability as a very low amount, like $100, or, at most, what you paid in the last year for the service.
Then you have clauses about uploaded content. Very often the content we upload can be used for marketing purposes or internal feedback and review. More often now, though, I see a lot of clauses about AI training – that your uploads or data can be used or sold to third parties to train AI models.
If tools like Termzy catch on, what do you think changes more – people’s behavior or the way companies write their policies?
Both? The goal is to change awareness and therefore to change behavior. You don’t scroll the internet without entering into multiple contracts every day and, right now, that happens almost completely unconsciously.
My hope is that more people at least see what they’re agreeing to and use that information to be more careful with the services they choose, especially in cases where they have an alternative.
This interview has been lightly edited and condensed.
🙏 *Thanks for reading! And since we’ve been talking about long privacy policies, it’s probably a good time to remember Longcat, the original longest cat on the internet. As always, feel free to drop me a line at allison@adexchanger.com with any comments or feedback. *