Are AI chatbots building the next walled garden?

Artificial intelligence chatbots have quickly become part of our everyday routines—from answering questions and drafting emails to planning trips and summarising long documents. But as these tools become more convenient, they also become more deeply embedded in our digital lives, quietly collecting and processing vast amounts of personal information in the background.

Many users don’t fully realise how much data is being gathered each time they open an AI chatbot app. Popular services such as Meta AI, Google Gemini, and ChatGPT routinely collect contact details, search and browsing history, and other user-generated content. According to new analysis by Surfshark, 70% of leading AI chatbot apps now also track users’ locations, a steep rise from 40% just a year earlier.

Don't miss: AI chatbots are getting 'aggressively' data hungry

By examining the privacy labels of the 10 most popular AI chatbot apps in the Apple App Store - and reviewing the privacy policies of DeepSeek and ChatGPT—Surfshark found significant differences in how aggressively these tools collect and link data. Meta AI stands out in particular, harvesting 33 of 35 possible data types and remaining the only app to collect information in the financial category. This raises pressing questions about what we trade away in exchange for AI-powered convenience - and how aware we really are of those trade-offs.

Another walled garden?

As AI chatbots ramp up data collection, a familiar pattern appears to be reemerging: the creation of another walled garden—this time built not from clicks, but from users’ thought processes.

In the social media era, walled gardens locked in behavioral data—likes, shares, and browsing patterns. With AI, what’s being locked in is cognitive data: users’ questions, reasoning, and vulnerabilities, said Jacky Chan, CTO of Votee AI and Beever AI.

"That's a qualitatively different kind of moat. When you confide in an AI assistant about a business decision or a health concern, that insight becomes platform advantage," he added.

As model capabilities commoditise, your conversations become the only real competitive differentiator these companies have.

For marketers in Asia–Pacific, this carries a specific edge, he explained. “Many brands here have already navigated the fragmented super-app ecosystem—WeChat, LINE, Grab, Gojek—each with its own walled data. AI adds another layer. If your customer data lives inside one AI vendor’s ecosystem and you can’t port it out, you’ve handed over leverage you’ll never get back.”

While that question looms, Dominique Rose Van-Winther, founder of Final Upgrade, said the more interesting story is the geographic divide in how people respond to this data collection.

"In Asia, there's a much higher comfort level. Historically, the dominant tools in those markets have operated with far fewer restrictions on what they gather. It's just been the norm. Europe sits at the opposite end, with GDPR setting the global benchmark for how seriously data protection is enforced."

That's the macro trend worth watching: organisations moving toward owning their full AI tech stack from the ground up rather than relying on third-party platforms where data flows are opaque, she said. "For marketers, this means choosing an AI tool is also a data jurisdiction decision. You're picking whose rules your customer data lives under. That looks very different depending on whether your customers are in Singapore, Berlin, or Texas, and it gets even more complicated when your AI vendor operates across all three."

Approaching ethical data use

To address this, the first step is actually understanding how AI agents operate, said Gary Liu, Terminal 3's co-founder and CEO. "When you integrate AI agents into marketing workflows, you quickly realise how efficient they are at completing tasks that pre-existing marketing tools and APIs cannot. What you may not fully appreciate is why they're so efficient — and the amount of private data they use to accomplish these goals."

"The way AI agents store and handle this data remains quite opaque to regulators, so it's very likely that what they're doing with data exists in a legal gray zone. AI agents generally have no built-in data privacy or security guardrails, and very few have ethical constraints on how they utilise data. If you are employing AI agents in your marketing workflow, it's incumbent on you to understand what kinds of private data they're collecting and how they're using it," he added.

Furthermore, brands need to stop treating data ethics as a compliance checkbox and start treating it as a design constraint, said Nathan Petralia, country head, Hong Kong, Ogilvy One and Verticurl. "At this point, the question isn't just what data you're allowed to collect, it's what data you actually need to deliver the outcome."

If you're running a CRM campaign and the AI tool you've plugged in is pulling location, health, and search history data you didn't ask for, that's a signal the integration wasn't scoped tightly enough.

Where it gets uncomfortable for brands is when marketers flip that lens inward, said Van-Winther. "What data are you collecting through the AI tools you've integrated? Do your customers know?"

Zoe Cheng, commercial director, APAC, Ekimetrics, said brands must adopt a strict purpose‑limitation principle, collect only what is strictly necessary to fulfil the user’s immediate request. "If a data point such as health history or precise location is not required for the task at hand, it should not be collected. Conversational data must be segregated from core marketing systems unless the user gives explicit, separate consent for future use."

Maintaining transparency and trust

Still, the perception that AI tools collect highly personal information is widespread—and varies by country, according to Kantar research. Location data is seen as the most commonly collected type across all age groups, especially among Boomers (67%) and Gen X (63%). While debate continues over whether AI chatbots are actually gathering personal data, perception becomes reality when it comes to trust. For brands, this means AI features must be introduced with transparency front and centre.

Petralia said transparency has to be operational, not just stated. "Publishing a privacy policy that nobody reads isn't transparency. What actually builds trust is giving users meaningful control, making opt-outs easy, and being honest about what the AI is doing with their data in plain language, not legal copy."

On the marketing front, the practical steps are: audit the data flows across every AI tool in your stack, make sure your consent architecture is current and reflects what you're actually collecting, and build a disclosure habit into your campaign briefs, he said.

Terminal 3's Liu said it is a huge assumption that consumers are actually becoming more aware of how much data their AI agents are collecting. If that awareness is the only motivation for marketers to act more ethically, we may be waiting a long time.

The better approach is to deploy AI agents in a technical environment that has inherent safeguards around data privacy, security, and agent governance.

Another key takeaway is consent fatigue, said Ekimetrics' Cheng. "We’re asking for more granular data than ever, but if the value we give back is just a slightly better retargeting ad, people will (and should) turn against it. If someone shares something sensitive with your AI, you owe them an immediate, tangible benefit something that feels like a concierge, not a data grab. If we don’t hold ourselves to that, regulators in this region will do it for us, and they won’t be gentle about it."

Mark your calendars for 24 June! #Content360 Hong Kong returns with a dynamic, one-day event dedicated to pivotal trends—from the silver economies to breakthrough IP collaborations, sports, and beyond. Let's dive into the art of curating content with creativity, critical thinking and confidence!

Related articles:

Chinese tech firms launch ChatGPT-style AI chatbots to the public
Why brands can't do without chatbots