---------------------------
Title: Your Calendar Knows You're Quitting
URL: https://usercentrics.com/knowledge-hub/inferred-data-ai-consent/
---------------------------

# Your Calendar Knows You're Quitting

As AI moves from generative to agentic, the traditional consent moment is disappearing. This piece argues privacy has to shift from governing what companies collect to governing what they’re capable of working out and why that starts with infrastructure not more consent pop-ups.

### At a Glance

- Historically, data privacy was more focused on what information you shared. AI introduces a new dilemma: what can be inferred from that data?
- 46 percent of consumers remain unclear about how their information is collected and processed.
- 60 percent say they’re uncomfortable with their data training AI models, but use the tools anyway.
- Agentic AI can act without asking first, removing the traditional moment of consent entirely.
- Solving this demands better infrastructure to regulate what AI systems can access, infer, and execute beforehand.

You never told your calendar app you were burning out. You never told your bank you were worried about money. And you never told your employer you were looking for another job. Yet, an AI system analysing your calendar activity, search history, or spending patterns might be able to infer all three.

That's what makes privacy more complex than simply deciding what information to share. Individual data points can reveal far more when connected, exposing patterns, relationships, and insights you never explicitly disclosed. The information you share is only part of the story. What can be inferred from it is another matter entirely.

For decades, one question has shaped how we think about privacy digitally: what did you give us permission to use?

A name.

An email address.

A browsing history.

A purchase record.

We built consent banners, privacy policies, and entire regulatory regimes around that exchange, as we should, since this kind of data needs protecting. However, protecting it was manageable because you knew what you’d handed over and could ask to see it again, edit it, or have it deleted.

Now we hand that data to AI and ask it to find patterns, draw conclusions, and predict what happens next.

Those conclusions are about real people, and most AI tools on the market give neither the company nor the person an easy way to see them, let alone question or correct them.

## What Happens to Information Nobody Ever Collected?

Picture an AI assistant at work. You’ve given it access to your calendar so it can schedule meetings, your email so it can summarize your inbox, and your company’s knowledge base so it can find documents.

None of those permissions feels unreasonable on its own.

In fact, in our [2026 State of Digital Trust research](https://usercentrics.com/resources/state-of-digital-trust-report-2026/), 49 percent of consumers are comfortable with AI assistants accessing their work tools, the highest comfort level of any category we tested.

- **49%** — comfortable with AI accessing work tools — the highest of any category
- **37%** — comfortable with access to financial accounts — the lowest

Now picture the assistant noticing a pattern. Your calendar has more meetings outside your department than usual. You’ve stopped joining a recurring call you used to attend every week. You’ve searched for compensation benchmarks and downloaded a handful of documents about a different industry.

Individually, none of that means much.

Together, it can look a lot like: **I’m planning to leave my job.**

## Are We Asking the Right Privacy Question?

Privacy teams have spent two decades on making sure they are able to answer this set of questions:

- what data do you have
- where did it come from
- how long do you keep it
- who can access it

And yet, our [2026 State of Digital Trust report](https://usercentrics.com/resources/state-of-digital-trust-report-2026/) found that 46 percent of consumers still don’t have a good understanding of how their data is collected and used. A figure that hasn’t moved in two years, and that’s even before anyone starts combining that collected data to draw new conclusions.

The understanding gap:

- **46%** — Global – flat for two years
- **56%** — Sweden
- **53%** — Germany
- **41%** — US
- **41%** — UK

A company can hold [thousands of ordinary data points](https://usercentrics.com/knowledge-hub/data-minimization/) about someone: purchases, searches, location, support tickets. And none of them reveals much on its own. Combine enough of them and it’s possible to draw out something new.

A prediction.

A preference.

An assumption that wasn’t present in any individual piece of data.

The old question was “what information do you have about me?”** AI adds a second one: **“what are you able to know about me?”**

These conclusions have a name: **inferred data**. EU data protection authorities have recognised this distinction since 2017, identifying three broad categories of personal data.

First, there's **data you provide**: your name, email address, or other details you knowingly share with a company.

Then there's **data companies observe**: the meetings you attend, the things you search for, the files you download, and how you interact with their services.

Finally, there's **data companies infer**: the conclusions they draw by connecting the first two. They might work out where you live based on your searches, whether you own a pet based on your shopping habits, or whether you're planning to leave your job based on your calendar activity and job searches.

You may never have explicitly shared any of this information. But that doesn't make it any less personal. Under the [General Data Protection Regulation (GDPR)](https://usercentrics.com/knowledge-hub/the-eu-general-data-protection-regulation/), inferred information can still qualify as personal data and receive the same protection. You have the right to see it and, in most cases, to object to the profiling that produced it.

But recognising inferred data as personal data is one thing. Accounting for it in practice is another.

So, do the AI tools we’re handing our data to have this distinction built in?

Mostly, no. Data protection law covers inferences, and a few of the biggest consumer platforms will now show you some of theirs. [Download your Spotify data](https://support.spotify.com/us/article/understanding-your-data/) and you’ll find a file called Inferences.json, listing the audience segments Spotify has assigned you from what you listen to and from what its advertising partners know about you.

LinkedIn does the same. Request the [full copy of your data](https://www.linkedin.com/help/linkedin/answer/a1339364/) from Settings, and the archive that arrives a day later includes a file called Inferences, holding the industry, seniority and interest categories LinkedIn has assigned to you from your profile and activity. It doesn’t say how it arrived at any of them.

But look at what’s in both: interest labels and advertising segments, the kind of inference built for selling ads, because that’s the kind of rules were written to catch.

The systems and processes companies rely on day to day haven’t caught up with the rest. Consent flows, data maps, access-request systems, even the checklists used in regulatory audits have always focused on what information was collected, not on what a system went on to conclude from it.

Which raises a bigger question: if an AI system has drawn conclusions about you, how do you find out what they are? And how do you get access to them?”

## AI Permissions Haven’t Caught Up With Consent

Consent is built around an exchange: a company explains what data it wants, why it wants it, and what choices you have. You can accept, refuse, or change your decision later. On websites, this process has become relatively standardised. Consent banners identify categories of data, explain their purposes, and record the choices people make.

But that model becomes harder to apply when the request is no longer about collecting a specific category of data, but granting an AI system access to an entire information environment.

Consider a seemingly straightforward prompt: **"Allow this AI assistant to access your email."**

What exactly does that permission cover? Reading individual messages? Summarising conversations? Identifying the people you communicate with most? Drawing conclusions about your interests, priorities, or intentions? Connecting information from your inbox with your calendar to build an even more detailed picture of your life?

The initial permission might be clear. Its potential implications are much harder to anticipate.

And that's where consent faces a different kind of challenge. It's no longer enough to explain what data a system can access. People also need a meaningful understanding of what that access could reveal, what the system might infer, and how those inferences could be used.

People have grown suspicious of AI, even if they can’t say exactly what’s bothering them.

Our 2026 State of Digital Trust report found that only 8 percent of consumers are fully comfortable with an AI assistant accessing their data without conditions.

Another 23 percent will allow it only if they can approve each request individually, and 17 percent are uncomfortable but allow it anyway.

The report describes this last group as giving [***resigned consent:***](https://usercentrics.com/resources/resigned-consent/) *agreeing to something despite feeling uneasy about it, often because opting out feels impractical, inconvenient, or like missing out on something they need. It's consent without genuine confidence or enthusiasm, and it raises an important question about how meaningful that agreement really is.*

- **8%** — fully comfortable
- **23%** — conditional
- **17%** — resigned
- **52%** — no view yet

Agentic AI pushes this further. Where a chatbot asks you to make a conscious choice about what to share, an agent can book, purchase, and share data on your behalf without asking each time.

In many of those cases, there’s no consent moment to point to at all. Our recent [report with MIT Technology Review Insights](https://usercentrics.com/resources/privacy-led-ux-ai-trust/) reframes the question for exactly this reason: **“who is consenting on behalf of the user, to what, and when?”**

Most organizations building agentic tools haven’t answered this question.

## Why Do We Keep Asking for the Very Thing That Creates the Problem?

People don’t want AI to know less.

They want it to remember their preferences, understand their schedule, and [anticipate what they’ll need next](https://usercentrics.com/knowledge-hub/hyper-personalization-vs-privacy-boundaries/). The more useful an AI assistant is, the more context it needs, and the more context it has, the more it can work out.

https://www.youtube.com/embed/3Y1sCjKK_8A?feature=oembed

Younger consumers already see where that leads. In our [Gen Z digital trust research](https://usercentrics.com/resources/winning-gen-z-marketers-guide-to-digital-trust/), Gen Z ranks AI assistant history among the data they most want protected, while older generations are more conscious of their login credentials and health records.

About 60 percent say they’re uncomfortable with their information being used to train AI models, a point higher than a year ago. And 52 percent now trust AI less than they trust humans with their personal data, both numbers are a few points higher than last year’s report.

And yet, more people use AI tools every year regardless.

Seven in ten find AI-driven personalization intrusive, yet among the consumers who actually read banners and understand their rights, more than half are comfortable with a company personalizing their experience. People aren’t rejecting the trade-off. They’re making it, often without a clear sense of what they’re trading.

2025

- **48%** — trust AI less than humans
- **59%** — uncomfortable with AI training

2026

- **52%** — trust AI less than humans
- **60%** — uncomfortable with AI training

Collecting less data wouldn’t fully solve this, since it would cost people the convenience they’re already asking for. Someone still has to decide where a system’s ability to work things out should stop, and who that someone is remains an open question for most organizations.

## What Happens When the System Gets It Wrong?

There's another layer to this problem: inferred data isn't necessarily accurate. AI systems work with probabilities, patterns, and correlations. They can identify signals that suggest something about you, but they can't always distinguish between what those signals actually mean and what they appear to mean.

Take the example of a system concluding that you're planning to leave your job based on your calendar activity and job searches.

Perhaps you're researching a competitor for a report. Perhaps you're helping a friend prepare for an interview. The same data can point to very different realities, but the system may settle on one interpretation and treat it as fact.

The consequences don't end with an inaccurate assumption. If that assumption informs a recommendation, influences an offer, or feeds into a decision about you, challenging it may be far from straightforward. You might not even know the assumption exists, let alone have an opportunity to correct it.

This points to a broader problem with meaningful consent: people are often expected to make informed decisions without having the time or context to understand what they're agreeing to.

Forrester's Enza Iannopollo made a related observation in our report with MIT Technology Review Insights, explaining why people often click through privacy notices:

> If you’re going to ask me 25 things in the first two seconds I’m on your website, chances are I’m going to skip through,” she says. “This isn’t because I don’t think privacy is important, but I’m there to accomplish a task, and reading the policy in-depth is not going to help me meet my goal.

The same tension becomes even more pronounced with AI. People are being asked to grant access to increasingly interconnected sources of information, while having little visibility into what systems might derive from that access.

And unlike a privacy notice, those inferences may not even be apparent at the moment someone gives permission.

## How Do We Govern Before the Agent Acts?

For organisations developing or adopting AI, this shift from data collection to data use introduces a different set of governance challenges. The question is no longer limited to what information a system can access, but also what it can do with that information, where it can send it, and which decisions it can influence.

With a generative AI chatbot, much of the governance challenge centres on transparency: explaining what information the system processes, how it is used, and what happens to it. These are important disclosures, but they become harder to manage as AI systems move beyond generating responses and begin taking actions.

Agentic AI introduces another level of complexity. An agent might retrieve information from one platform, pass it to another, or initiate an action based on what it has learned, potentially without the user being aware of each step. By the time someone discovers how their information has been used, the data may already have moved beyond its original context. This makes it essential to establish permissions and boundaries before an agent acts, rather than relying solely on transparency after the event.

The technical infrastructure for this level of governance is still evolving. The Model Context Protocol (MCP) is one emerging part of that picture. It provides a standardised way for AI applications to connect with external tools and data sources. Combined with an appropriate policy and consent layer, this could allow organisations to define what agents can access, monitor and log their actions, and carry existing user preferences across different interactions and services.

When our report with MIT Technology Review Insights examined this in April, MCP was less than a year old, and most organizations deploying agentic tools hadn’t built anything resembling a policy layer for it.

That’s part of why bringing that capability in-house became a priority when we [acquired MCP Manager](https://usercentrics.com/knowledge-hub/welcoming-mcp-manager-governing-consent-in-ai-workflows/). The consent someone gives (or doesn’t) at one point shouldn’t stop applying once their data moves somewhere else, including into the agents acting on their behalf, and that’s also the thinking behind how we’ve approached recent shifts like the [EU AI Act’s transparency and consent requirements](https://usercentrics.com/knowledge-hub/eu-ai-act-high-risk-delay-article-50-transparency-consent/).

Consumers are already pointing to where that line sits. In the 2026 research, 65 percent still want a human for healthcare interactions, and 61 percent for financial advice and for handling complaints. Those are not the places to keep AI out. They are the places where the permission layer has to be in place first.

**Human preference by context (2026 report):**

- **65%** — healthcare
- **61%** — financial advice
- **61%** — complaints

## Which Question Haven’t We Finished Answering?

We’ve spent 20 years answering one question: what can a company do with the data you give it? Consent banners, privacy policies, and the right to say no all came out of that work.

AI adds a second question: what can a company infer about you? It needs the same answer. Decide where useful prediction ends and unwanted knowledge begins, then extend the consent infrastructure we already have to hold that line before an AI system acts, not after.

We built privacy around “what did you tell us.” The next phase builds on it, around “what did we conclude,” and what we chose not to.

*Usercentrics does not provide legal advice, and information is provided for educational purposes only. We always recommend engaging qualified legal counsel or privacy specialists regarding data privacy and protection issues and operations.*

## See What Consumers Expect Before AI Gets Access

The State of Digital Trust 2026 surveys 11,000 consumers across seven markets on AI, consent, and what they will and won’t allow an AI assistant to reach.

---

## Footer

### Products
- [Usercentrics Web CMP](https://usercentrics.com/website-consent-management/)
- [Usercentrics App CMP](https://usercentrics.com/in-app-sdk/)
- [Usercentrics CTV CMP](https://usercentrics.com/usercentrics-ctv-cmp/)
- [Usercentrics Privacy Policy Generator](https://usercentrics.com/privacy-policy-generator/)
- [Server Google Tag Manager (sGTM)](https://usercentrics.com/server-side-tracking-solution/)
- [Meta signals gateway](https://usercentrics.com/meta-signals-gateway/)
- [MCP Manager by Usercentrics](https://usercentrics.com/ai-model-context-protocol-manager/)
- [Web compliance scan](https://usercentrics.com/privacy-compliance-scanner/)
- [App compliance scan](https://usercentrics.com/app-data-privacy-audit/)
- [CMS Integrations](https://usercentrics.com/integrations/)

### Solutions
- [Data Privacy Regulatory Compliance](https://usercentrics.com/data-privacy-regulatory-compliance/)
- [Marketing Performance Optimization](https://usercentrics.com/marketing-performance-optimization/)
- [Migration](https://usercentrics.com/migration/)
- [Ad monetization](https://usercentrics.com/ad-monetization/)
- [First-Party Data Collection](https://usercentrics.com/server-side-tracking/)
- [Media & Publishing](https://usercentrics.com/media-publishing/)
- [Retail &amp; E-commerce](https://usercentrics.com/retail-ecommerce/)
- [Banking, Finance &amp; Insurance](https://usercentrics.com/banking-finance-insurance/)
- [Healthcare & Pharmaceuticals](https://usercentrics.com/healthcare-pharmaceuticals/)
- [Gaming](https://usercentrics.com/gaming/)
- [Education](https://usercentrics.com/education/)
- [Automotive](https://usercentrics.com/automotive/)
- [Travel & Hospitality](https://usercentrics.com/travel/)

### Regulations
- [GDPR (EU)](https://usercentrics.com/gdpr/)
- [GDPR (UK)](https://usercentrics.com/uk-gdpr/)
- [CCPA (California)](https://usercentrics.com/ccpa/)
- [TCF v2.4 (IAB)](https://usercentrics.com/cmp-for-publishers/)
- [DMA (EU)](https://usercentrics.com/digital-markets-act-dma/)
- [Amazon Consent Signal](https://usercentrics.com/usercentrics-cmp-and-amazon-consent-signal/)
- [Google Consent Mode (EU)](https://usercentrics.com/usercentrics-cmp-and-google-consent-mode-v2/)
- [Microsoft UET Consent Mode (EU)](https://usercentrics.com/usercentrics-cmp-and-microsoft-consent-mode/)
- [Microsoft Clarity Consent Mode](https://usercentrics.com/usercentrics-cmp-and-microsoft-clarity-consent-mode/)
- [All regulations](https://usercentrics.com/regulations-and-frameworks/)

### Resources
- [Blog](https://usercentrics.com/knowledge-hub/)
- [Guides](https://usercentrics.com/guides/)
- [Checklists](https://usercentrics.com/checklists/)
- [Whitepapers](https://usercentrics.com/whitepapers/)
- [Case studies](https://usercentrics.com/customer-stories/)
- [Magazine](https://usercentrics.com/magazine/)
- [Events](https://usercentrics.com/webinar/)
- [TV Show](https://usercentrics.com/privacy-led-marketing-show/)
- [State of Digital Trust 2026](https://usercentrics.com/resources/state-of-digital-trust-report-2026/)
- [Privacy-Led Marketing](https://usercentrics.com/privacy-led-marketing/)
- [Courses](https://courses.usercentrics.com/)
- [RFI template](https://usercentrics.com/resources/usercentrics-rfi-template/)
- [ROAS Calculator](https://usercentrics.com/roas-calculator/)
- [Ad revenue calculator](https://usercentrics.com/ad-revenue-loss-calculator/)
- [Developer documentation](https://usercentrics.com/docs/)
- [Release notes](https://releases.usercentrics.com/en)
- [Support](https://support.usercentrics.com/hc/en-us)
- [All resources](https://usercentrics.com/resources/)

### Company
- [About us](https://usercentrics.com/about-us/)
- [Press](https://usercentrics.com/press/)
- [Our offices](https://usercentrics.com/contact/)
- [Careers](https://usercentrics.com/career/)
- [Diversity and inclusion](https://usercentrics.com/dei/)
- [Trust center](https://trust.usercentrics.com/)
- [Open positions](https://apply.workable.com/usercentrics/)

### Support
- [General support](https://support.usercentrics.com/hc/en-us)
- [Contact sales](https://usercentrics.com/book-a-consultation/)
- [Technical support](https://support.usercentrics.com/hc/en-us/requests/new)
- [Suggest a feature](https://support.usercentrics.com/hc/en-us/requests/new?ticket_form_id=10610312381340)
- [Billing and account](https://support.usercentrics.com/hc/en-us/categories/12253804608156-Account-and-billing)
- [Partner login](https://partnerportal.usercentrics.com/)
- [Partner program](https://usercentrics.com/partner-program-overview/)
- [Affiliate program](https://usercentrics.com/affiliates/)