As businesses increasingly adopt artificial intelligence, understanding how AI interacts with data privacy laws is essential for managing risks and protecting personal information.
In this session from Termly’s 2026 Partner Summit, privacy attorney Anokhy Desai explores the intersection of AI and data privacy, including requirements under the GDPR, CCPA, and EU AI Act.
Learn practical strategies for evaluating AI vendors, managing consent and consumer rights, improving transparency, protecting sensitive data, and building responsible AI governance practices.
Read the transcript
0:04
So good morning for the West Coast, afternoon for the rest of us, wherever you're tuning in from, to this presentation on navigating data privacy in an
0:13
AI-driven world. I'm Anokhy, and I'll be guiding you through how businesses can leverage AI responsibly while staying compliant with global privacy laws.
0:23
Over the next 45 minutes or so, we'll cover key concepts, challenges, and actionable best practices.
0:31
And in our roadmap today, I'll be introducing myself a little more fully, and then we'll review some key terms for anyone new to this practice area,
0:40
look at how AI and privacy intersect, go over some common compliance challenges and best practices for businesses based on those challenges, and we'll keep this interactive.
0:51
So feel free to submit questions throughout.
0:53
Also, that hashtag to get that swag afterwards, and we'll get to the questions at the end.
0:59
This session will be recorded, and the deck will be available for download after, along with the resources and checklists that I'll mention in the presentation.
1:07
And as always, legal disclaimer here.
1:10
The presentation and any information that I'll be sharing today from the wonderful folks at Termly is for
1:16
informational purposes only and shouldn't be construed as legal advice from me or my employer.
1:22
But if you do have a more complex question that you think can't be answered here today, do feel free to email me, and I'll put my email in the chat towards the end.
1:34
So, my name is Anokhy Desai.
1:36
I'm with the law firm Husch Blackwell in the Privacy, Cyber, and AI group and have previously worked with the IAPP, with startups in-house as a technologist, a handful of roles.
1:48
I have a bunch of privacy certs, and I'm in the process of studying for the AIGP certification, and have worked with cybersecurity, privacy, and AI issues for businesses of varying sizes.
2:01
As the term is used today, AI, artificial intelligence, refers to a lot of different systems.
2:07
Some mimic human language, learning, and decision-making without much human intervention.
2:13
For businesses, AI is becoming mainstream at a rapid pace.
2:17
According to recent findings by Termly, 77% of companies are already exploring or using AI tools.
2:24
And in 2024, there was reportedly over a 100% increase in generative AI adoption.
2:31
If you're here, chances are that your business already uses one or more of these AI platforms for various activities.
2:38
So feel free to share which one of those you're using in the session chat, or any relevant details about your experiences when it comes to compliance.
2:46
But in the meantime, let's break down the types of AI that are most relevant to businesses.
2:52
Machine learning powers predictive analytics, which helps with forecasting trends and customer behavior.
2:58
Natural language processing is used in chatbots and sentiment analysis for better customer engagement.
3:05
Computer vision uses deep learning and neural networks for object recognition in processes like inventory management and retail and facial recognition in security.
3:16
And generative AI is the one I'm sure we're most familiar with here.
3:20
That includes ChatGPT and Claude, a bunch of others, and as the name indicates, generates content at scale.
3:27
Actually, some of those, including ChatGPT, now have integrated computer vision.
3:32
And so you can take a picture and ask it to describe what that's a combination of different types of AI.
3:39
There are so many instances of AI transforming day-to-day operations like
3:46
customer support, automation, fraud detection, personalized marketing, and even supply chain optimization.
3:53
But the catch is, to function properly, AI relies on a lot of data, like literally millions of pieces of data to function better.
4:02
And sometimes that data falls under the legal definition of personal information or personal data.
4:08
And this means that businesses do have to consider privacy compliance from the beginning.
4:13
Whenever you choose an AI platform for your business, using AI without considering privacy laws, common
4:22
ones are GDPR and CCPA, can have significant consequences, but we'll cover that more later.
4:28
But some examples include hefty fines, reputational damage, and significant loss of consumer trust, which we've
4:34
already seen in 2025 alone has had a real impact on the earnings of major organizations.
4:41
As we move forward, keep in mind that choosing to use AI isn't just a matter of upgrading the team's tech stack, but also a compliance challenge and an opportunity to build trust through responsible use.
5:00
So data privacy is the principle that individuals should have control over their personal information and surrounding processes like how it's collected,
5:08
how it's used, how it's stored, and even whether it can be deleted.
5:12
It's not just a legal requirement, but also a trust factor for consumers, as highlighted by recent privacy stats.
5:19
Like a study from Cisco recently showed that 75% of consumers will not purchase from organizations that they don't trust with their data.
5:29
We're seeing retail organizations face this issue right now.
5:34
Some core principles of data privacy are data minimization, which is collecting only what you need; purpose limitation, which is using data only for
5:42
the purpose that you disclosed; transparency, which is clearly communicating your data practices; and security, which,
5:50
in conjunction with privacy, means implementing appropriate safeguards to prevent security incidents like breaches.
5:57
More and more, we see security and privacy being tied in together when it comes to compliance needs, as we can see by the new 2026 CCPA
6:06
regulations, which require cyber audits on top of the regular privacy risk audits for certain organizations.
6:14
So these core privacy principles, data minimization, purpose limitation, transparency, and security, are enforced through privacy laws globally, just to name a few.
6:24
There's GDPR in the EU, CCPA in California, the LGPD in Brazil, POPIA in South Africa, the PDPA
6:33
in Singapore, and a bunch more that all require businesses to respect user rights first and have business obligations for complying with certain rules.
6:44
These rights, the consumer rights, typically include the user's ability to access, correct, delete, and port their data to other services.
6:53
For businesses, complying with these laws means providing clear privacy notices, implementing consent management for sensitive
7:00
data, and the consent management should be plugged in correctly in the back, and ensuring users can exercise their rights easily.
7:08
Aside from meeting compliance requirements and avoiding fines,
7:11
this means respecting user privacy is shown to build trust and loyalty with your consumers.
7:16
Termly research actually shows that 94% of businesses believe it's possible to balance marketing goals with privacy compliance.
7:23
And that's something we'll continue to talk about today. I see a Termly shout-out in
7:32
the chat, using Termly's Policy Generator to update the privacy policy to include AI.
7:37
Always good to hear that people are updating their policies to match their existing practices.
7:44
So most consumer-use AI today requires data centers and troves of data to function.
7:50
Like I mentioned earlier, some of this data can fall under the legal definition of personal information or personal data, which is protected by global privacy laws.
7:59
Sometimes the data shared with these AI platforms can even include sensitive information, which is a special category of data for anyone new to privacy
8:08
here that requires even more protection by businesses.
8:11
If this data is mishandled, the business risks fines, lawsuits, and, of course, reputational damage.
8:19
By now, it's clear that AI and privacy are not separate, but rather very intertwined.
8:24
And every time a business uses AI, it's relying on data, which is why they're intertwined.
8:30
It relies on names, emails, consumer behavioral patterns, and even sensitive health or financial information.
8:38
These are the same categories of information that privacy laws like the GDPR and CCPA are designed to protect.
8:46
For businesses, this has led to an unavoidable dilemma. AI promises efficiency, personalization, and innovation.
8:54
But it also introduces this risk.
8:56
If businesses fail to handle personal data responsibly, the consequences can be rough for the consumer whose privacy is at risk, on top of the existing risk for the business.
9:08
Under the GDPR, we know fines can reach 20 million euros or 4% of annual global turnover, whichever is higher.
9:16
But that financial penalty is only part of the story.
9:19
The rest of the cost is the loss of consumer trust and damaged reputation.
9:24
Consumers today are privacy-conscious, and we keep hearing this over the years. It's only more and more true as the years pass.
9:32
And one misstep can be costly.
9:34
This means transparency and consent are no longer optional.
9:38
But if you're watching this from an EU member country or you're already GDPR compliant, you may already be doing this.
9:46
But if you're using AI tools, you should also be clearly informing the users about how their data is being used.
9:52
And when necessary, you should be getting explicit consent for things like sensitive information.
9:59
It is equally important to ensure that consumers can exercise their privacy rights as much as you are saying that they can in your privacy policies, whether
10:07
that's accessing, correcting, or deleting their data, even if it was processed by an AI system.
10:15
So for high-risk applications like facial recognition or consumer profiling, DPIAs, for folks new to privacy, that's Data Protection
10:24
Impact Assessments, are also essential.
10:27
These act as a sort of safeguard against potential ethical, security, and legal pitfalls.
10:32
The bottom line is that AI and data privacy are inseparable, and responsible AI use is not just a technical challenge, but also a governance one.
10:41
Businesses that embrace this reality can not only avoid penalties, but also build trust and gain that extra bit of competitive advantage in this AI world.
10:58
Moving on to data privacy laws in a little more detail, nearly every major privacy regulation
11:04
now intersects with AI, and businesses need to understand what that means in practice.
11:11
So I'll start with the EU AI Act, which is, I believe, the first comprehensive framework for AI governance in the world.
11:19
It introduces a risk-based approach where AI systems are categorized as minimal, limited, high-risk, or prohibited.
11:27
High-risk systems like those that are used for facial recognition or credit scoring face the strictest requirements, which is unsurprising,
11:36
which includes obligations like transparency, documentation, and ongoing risk assessments.
11:42
And it applies broadly to anyone making, using, importing, or distributing AI systems in EU member states.
11:50
So if you're based in the U.S., you may still fall under the EU AI Act.
11:55
If your business is using AI in this way, compliance with the Act is probably mandatory.
12:02
Then there's everyone's favorite GDPR, which remains the contentious gold standard for data protection.
12:09
I'm not sure if anyone saw NOYB's post about their love-hate relationship with GDPR, but I know privacy pros feel a similar way as well.
12:18
We'll see if anything changes, or if anything else, new updates are adopted after the Omnibus package goes through trilogue.
12:25
But for now, under the GDPR, if you're processing personal data, whether it's through an AI system or otherwise, you need to have one
12:33
of six lawful bases: consent, contractual necessity, legal obligation, vital interests, public task, or legitimate interest.
12:42
Although the last one has become more limited over the years.
12:48
It's not just enough to say that we're using AI for efficiency, and this is something that we should really talk to stakeholders within our companies about,
12:56
but you need to justify why you're collecting data and how you're using it, and make sure that individuals can exercise their rights.
13:05
So that means if a customer asks to delete their data, you should be able to remove it, which is something to keep in mind before that data becomes embedded in an AI model.
13:16
In the U.S., the CCPA adds another layer where businesses must provide clear privacy notices, allow consumers to
13:23
opt out of the selling or sharing of their personal data, and obtain opt-in consent for sensitive data like health or financial information.
13:31
On top of that, with the new CCPA 2026 regulations, you need to make sure that the consumer choice that they choose to opt into or out of is symmetrical.
13:42
So there needs to be choice symmetry and no dark patterns there.
13:49
If you use an AI system with this type of data, you may need explicit permission from applicable consumers. I'm seeing, what about,
13:58
yes, thank you, Canada. What about PIPEDA and Quebec Law 25? I can get back to you on that via email.
14:06
I'm not actually sure if AI regulations have been advanced in both areas. But Jeff, we can talk about that offline.
14:15
Even though the legal landscape is very much in development, AI itself doesn't exist in a legal vacuum.
14:22
Every privacy law you already know, the GDPR, CCPA, LGPD, POPIA, PIPEDA, for our Canadians out there, can apply to AI-driven data processing.
14:36
And the penalties for getting it wrong are steep.
14:38
We know the 20 million euros or 4% of global turnover for GDPR. We know the $7,500 per violation for CCPA.
14:46
But again, beyond the fines, the headlines can actually cause a real dent.
14:51
Consumers expect transparency and accountability, and regulators are watching closely.
14:57
So as you innovate with AI, as your company pushes for AI innovation within the organization, ask yourself, are we compliant?
15:05
Are we being transparent? Are we prepared to honor user rights?
15:10
And does our privacy policy reflect our actual processes?
15:15
As we're surrounded by the push for AI use, we should prioritize both legal compliance and keeping an eye on consumer trust.
15:22
But when businesses adopt AI, the excitement can overshadow the complexity. And I do understand this.
15:28
Everyone gets excited about a new tool, especially when it promises to deliver higher efficiency.
15:35
But behind the promises of efficiency and innovation lies a set of challenges that can make or break compliance and consumer trust.
15:44
The first challenge is identifying responsible AI tools. Not all AI platforms are created equal.
15:50
Some lack transparency about how they process data or to what extent they comply with privacy laws.
15:56
And it's not always, you know, on them for missing this.
16:00
Sometimes they will develop faster than their privacy policies and terms of use can catch up with.
16:07
But choosing the wrong vendor for your business can expose you to legal risk, not just the vendor.
16:14
Responsible selection means reading privacy policies carefully, checking for certifications like ISO 27001 or SOC 2, and
16:21
ensuring that the vendor aligns with frameworks like the EU AI Act or, more popular in the U.S., the NIST AI Risk Management Framework.
16:31
And the next challenge is managing user consent, preferences, and rights.
16:35
AI systems often process personal data in ways that aren't immediately visible to consumers.
16:42
So if you're using AI for personalization or analytics, you're going to need robust consent management processes.
16:48
This includes clear banners, preference centers, and mechanisms for users to opt in or out.
16:55
And under privacy laws like the GDPR and CCPA, users do have the right to access, correct, and delete their data, even if it's embedded in the AI model.
17:05
I know privacy folks are tired of hearing about cookie banners as much as users are tired of navigating through them.
17:12
So try to make sure that the language and UI are user-friendly.
17:17
Another major challenge here is preventing unauthorized data sharing or access.
17:23
AI systems can inadvertently expose sensitive data if security measures are not strong enough.
17:29
So this means implementing stricter access controls, encryption, and regular audits.
17:35
It also means training your team because human error is often the weakest link in data protection.
17:40
I know this from just having worked in cyber or privacy. I'm sure many of you watching can relate to this.
17:48
Let me know, actually, in the chat if you were just as busy as me after the holidays because of security incidents from human error.
17:57
Finally, there's ensuring transparency with consumers.
18:01
People want to know when AI is being used and how it affects them.
18:06
If your chatbot is powered by AI, then there are laws that require you to say that it is.
18:11
If you're using AI for automated decision-making, privacy laws require you to disclose that. Transparency isn't just about good ethics.
18:20
It's a legal requirement under existing and emerging regulations.
18:23
These challenges aren't insurmountable, though, but they do require planning, governance, and a commitment to privacy-first innovation.
18:34
I see someone in the chat. Biggest human error is silly, obvious passwords. It's silly to some and unique to others.
18:41
I really do think those really long string passwords do become more effective over time.
18:47
If folks have a whole sentence in mind, we'd rather they use that than, you know, fluffy two six one.
18:59
All right, okay, this is a slide that I know can really come in handy, and this checklist will be available afterwards.
19:07
So as much as we say not to treat compliance as a checklist, it is nice to have this starting place if you're building out AI governance from scratch.
19:16
Termly was kind enough to make this 10-step checklist, and yes, it will be available afterwards.
19:23
First step: selecting privacy-compliant AI tools. That is easier said than done.
19:29
We recommend that you'll want to verify their compliance with privacy laws that are applicable to you and your organization.
19:36
Also, what jurisdictions that you're in, checking their certifications and cybersecurity practices, and confirming vendor policies do match their practices.
19:48
Sorry.
19:48
That vendor policies do match their practices when it comes to the data lifecycle.
19:53
Step two is a privacy lawyer's bread and butter: terms of service, privacy policy, liability review.
20:00
Making sure that the vendor doesn't use data for model training without consent, and if they're going to be using data for purposes other than what was originally stated, that
20:10
you do address that in a new agreement or that they do run a new risk assessment on that and notify you so that you can obtain user consent.
20:20
A third step is data limitation: only sharing data that's necessary with AI.
20:27
Just because you're typing and not on the phone doesn't mean that those other rules go out the window.
20:32
So don't share sensitive information like your mother's maiden name, your place of birth, and within your organization,
20:38
make sure that there are data minimization policies in place and that employees know what that means.
20:46
I often see organizations say that we have our terms of use in place.
20:50
We do have a privacy policy, a separate cookie policy, a separate AI use policy, but employees often don't review them or
20:59
it's a checkbox, "I have read," kind of situation, and there's not an existing training in place to explain why we have
21:06
all these things and what that means for them and what actions they should be taking, if any.
21:15
So steps four through six, standard practice for privacy people.
21:19
Make sure to remove PII before processing data or anonymize the data before using it to train AI models.
21:27
Also, preferring aggregated datasets for analytics whenever possible over pseudonymized. Step five: using personal data.
21:38
Using personal data always means informing users and obtaining explicit consent with opt-in consent for sensitive data like health and financial information.
21:47
Just making sure that users are informed about what happens to their data.
21:50
Because the biggest thing here, especially if you're a company that often gets consumer complaints or that faces any kind
21:57
of U.S.-based privacy litigation, it stems from a lack of transparency and consent to the consumer.
22:04
So if they don't understand what's happening, you just want to make sure that they see it in plain language.
22:12
Step six, do make sure that your opt-out mechanisms work.
22:17
And I know that's a bit of a technological challenge depending on what your company is doing with the data, how big your company is, how deep
22:25
your technical knowledge is, so I'm not going to try to generalize a piece of advice for this other than making sure that you have the right people in the room for business conversations about AI.
22:40
Next, you'll want to audit your company's AI usage through activity logs, data flows, and regular risk assessments.
22:48
Documentation is key.
22:50
If you do get that knock on the door from another lawyer, from an enforcement agency.
22:56
And on the cybersecurity side, role-based access controls are a good place to start, along with encrypting data at rest and in transit.
23:05
If you're using an AI vendor, make sure to talk to them to see what their update process is like, whether it's automated updates
23:12
pushed through or monthly planned occurrences that you'll need to notify relevant employees about.
23:19
And you'll also want to prepare for breaches, update your incident response, disaster recovery, business continuity plans to account for your company's AI tools, if relevant.
23:30
But back to that vendor side, when you're reviewing agreements and policies, take note of their breach notification obligations and warranties.
23:42
On the incident response thing, one last note.
23:44
I have seen organizations have incident response plans in place, appropriate staff on their security teams
23:53
and IT staff notified on how to place data holds, that kind of stuff.
23:59
But these plans often don't actually get walked through.
24:02
And so when it comes to actually following through what the plan is, I've seen organizations not know
24:09
where certain things are or not know who to contact when it comes to, you know, quarantining data.
24:16
So do make sure that you walk through the plan, you have a simulation. Last, final step: staying ahead of regulations.
24:28
Also easier said than done. There are a lot of resources out there.
24:31
Termly produces resources as one, a lot of folks like myself producing blog posts to just make sure that folks stay apprised of all
24:41
of the upcoming regulations, especially with draft regulations constantly being updated, with things like 2026 CCPA updates that came into effect this January 1st.
24:55
There are some assumptions even I make about different types of generative AI users, where I'll assume that
25:02
if you work in technology, you're aware that your ChatGPT prompts can be made public.
25:08
But that's not the case, as we've seen with various headlines and research studies that have exposed ChatGPT conversations in whole and prompts.
25:20
So training is important, making sure that people understand how much data they can enter into,
25:28
let's say, generative AI, how much data is being used to train models.
25:33
Training staff on these emerging compliance requirements and ethical AI practices can actually go a long way if they understand it.
25:46
Great. Q&A, Q&A time.
25:49
I know for some folks this is your lunch break, and so you may need a little fuel before you come up with questions again.
25:56
I will put my email down here in the chat if you have a more complex question or want to reach out separately.
26:12
All right, we're marketing. We deployed an AI chatbot through our CRM website. For customer service, we deployed an AI chatbot.
26:22
All right, more of a comment than a question.
26:24
But I actually will touch on this. I have worked with clients who've had questions about privacy compliance when it comes to their AI chatbots.
26:34
The biggest thing here is transparency and disclosures.
26:38
So when it came to at least their compliance, and this may be similar for folks watching who are in
26:44
a similar place, they wanted to make sure that they didn't have to be litigation-ready and that they were compliant from the beginning.
26:55
So in the states that they were located, there were rules about, and also U.S.-based best practices
27:04
about disclosing that the chatbot is a bot, that it is generative AI and not a human person.
27:13
So when you use generative AI for chatbot reasons, make
27:18
sure that it does say, you know, chatbot as the user, or that there is a notice when, like, when
27:28
the user clicks on the chatbot that has a pop-up of, like, "Have any questions? Want to ask us some questions? We'll be happy to answer them."
27:35
When the user clicks on that dialog box, that there is a notice that says, you know, "This communication is with Company ABC's chatbot."
27:48
The communications may be recorded or stored for whatever purposes, internal purposes.
27:55
And then you can go ahead with the, you know, "How can we help?" Some platforms have character limits.
28:01
So that is a creative challenge that we worked on, on how to get all of that information across in a limited number of characters and words.
28:12
But yeah, just when it comes to using AI chatbots and privacy compliance, a big way to
28:20
stay compliant there is to make sure that the user knows it's a bot and not a human.
28:26
Even if the conversation does get referred to a human, a customer service rep later, that they know that first part is with a
28:33
bot, and that their conversation may be stored within company logs for whatever purposes.
28:43
So that they can opt out of using that chatbot feature if they don't consent to that. Right.
28:53
Go through another question here.
28:57
Do you have any suggestions for what transparency is regarding the use of AI for our business?
29:03
Is there something that should be mentioned in a privacy policy or other legal document?
29:10
Yeah, actually exactly that, in your privacy policy.
29:14
Now I did mention that some folks have a different privacy policy from their AI use policy. As far as I know, and this may have changed,
29:23
we're in a new calendar year, that's not required.
29:27
But you can, it may be a little more clear depending on how lengthy your privacy policy is
29:33
to have a different document for AI compliance and just a place to note your AI use and practices and what data is collected there.
29:44
It may be redundant to the privacy policy depending on how many tools or how few tools you use.
29:51
But what I typically have been doing is adding a
29:57
clause within the privacy policy that lists the following tools
30:04
use AI, and then listing out kind of a summary of what the tool does, what type of data it
30:11
uses or collects, if it's consumer data or even B2B data, what the purpose of that data collection and processing is, and how long that data is stored.
30:22
Now sometimes it's instantaneous, like with facial recognition technology, but even then that has a little more stringent laws in place.
30:31
With biometric laws across the country, and with more state-level biometric laws popping up every day, keeping an eye on what passes there.
30:44
Even with things like banks that use voiceprints for fraud purposes, I know that they have,
30:52
depending on the size of the financial institution, they have a separate biometric data privacy policy in place
31:01
and that will include the use of AI for that kind of purpose, typically for fraud purposes.
31:07
As long as you're transparent, you're getting user consent for this, and you're letting users know that the data, the conversation, their voice may be stored for however long for security purposes.
31:20
That is above board when it comes to consent. But I think I missed one part of that question. What does transparency look like?
31:29
Yeah, transparency. So it's one of two things.
31:32
If it's something like a chatbot, it's making sure that in that quick interaction, it's clear to someone outside of our field, it's clear to a layperson, to the consumer, that it is AI.
31:45
The biggest thing here is you don't want to be tricking anyone.
31:48
This falls into that dark patterns category, and that can get gray very quickly.
31:52
But as long as you're making it clear that it is a chatbot, that it is generative AI, that it is an LLM and not a human, that counts as disclosure and transparency.
32:02
And then separately in a privacy policy or AI use policy, making sure that what you're
32:09
documenting there matches your practices and also matches the tools that you're using and what they say they're doing, that's really important.
32:17
So getting ahead of that by discussing with the vendor what they actually do and say and making sure that that matches their privacy policies as well is really important.
32:29
All right, what about risk factors for AI agents?
32:33
For instance, an AI voice receptionist's risk profile is low compared to an AI recruitment agent
32:40
which is ranking candidates, which could be a higher risk as drift and bias become factors.
32:47
Do we need to disclose this?
32:50
Yeah, any AI use, especially with agentic AI becoming more popular nowadays, it's tough because a lot of
33:01
agentic AI is either for individual consumer use, and so they will see the appropriate disclosures necessary in the terms of use or in the privacy policy.
33:11
Although, you know, this is a study that comes up time and time again for these Data Privacy Day conferences.
33:17
But you know, how many consumers are actually reading these privacy policies?
33:22
How many privacy people are actually reading these privacy policies? I think putting it in the policy may not be enough.
33:29
Especially when it comes to tools that are heavy on data collection and processing and data access, that
33:37
it might be a good idea to discuss with the appropriate folks in your company.
33:44
If you develop agentic AI, to have just-in-time notices or pop-ups, kind of like whenever there's a new feature
33:52
in a Google Suite product, in your Gmail, they'll have a little pop-up that says, "Walk me through it." And some of those are not optional.
34:01
You can't just say, like, X out of it.
34:03
It will say, like, you will have to click next, next, next, but it'll show on the various points of the application where the changes were made.
34:11
That is something that I've seen some apps try to integrate when it comes to new feature design.
34:16
So that users are aware because they know that the users are not reading the privacy policy or the new updates in the App Store when they download an update to a mobile app.
34:28
But what I was saying was there are two users, two main users of agentic AI.
34:32
There's the consumer-use one where consumers want a personal admin.
34:37
You know, I hate calling my dentist, but it would be less of a pain if I could automate that and have an AI agent do that for me while I'm delivering a presentation.
34:50
And then there's also the internal business use where, you know, a partner-level person at a firm or even a team
35:00
who needs an extra admin member would use an AI agent to kind of do meeting note wrap-ups for them or
35:08
you know, calendar invites, sending a summary message to folks that attended a meeting after the fact.
35:17
And all of those have different purposes and use and collect different levels of data.
35:24
That can be tougher because as organizations buy enterprise-level agentic AI solutions, it becomes even less likely
35:33
that the employees using this have truly read or understood the terms of use or privacy policy there.
35:41
So for that kind of use, I would really push the just-in-time notice type of thing.
35:45
The pop-ups, as it becomes relevant on the different parts of how to use this, or kind of a mandatory tutorial on, like, this is how this agent sets up a calendar
35:57
invite. The data this uses: you know, the people listed in prior emails, suggested contacts from your
36:05
contact list that you frequently communicate with in conjunction with the people in the last email, the frequency of meetings you have with these people.
36:14
So we have access to your calendar as well.
36:16
So having that kind of notice so users understand how it works.
36:20
Because all of that data collection can be necessary for these tools to actually function correctly.
36:27
But we want to watch out for that creep factor, which is a big thing in privacy.
36:30
You want to be able to disclose how much data is being collected and processed to consumers, but not so much that you freak them out from using the actual tool that they wanted to use in the first place.
36:43
It is a tough line to balance.
36:48
All right, take a little water break and then read the other questions. Okay.
36:57
Does my U.S.-based business need to worry about the EU AI Act? So yes, it's kind of a GDPR situation here.
37:05
If you are a business that fits in the categories of using, deploying AI, even if you're U.S.-
37:13
based, if you deploy AI within an EU member state and meet these other categories, you would fall under EU AI Act compliance.
37:24
Now one thing about AI laws is they kind of span the gamut, but similar to privacy laws, they require a lot of the same base core principles of transparency, consent, data minimization.
37:38
So really trying to get ahead of something like a huge breach where so much data has been ingested that wasn't actually needed and now we have to worry about the output of all of this.
37:57
Do data privacy laws impact how my business uses AI?
38:02
Maybe, sorry for the legal answer, but it does depend. If your business uses, you know, a help
38:12
desk chatbot that tries to categorize an incoming client's question so it can be sent to the right
38:20
person internal to the company, or lead them to an FAQ page elsewhere on your website, or even provide other helpful links on your website.
38:30
The privacy considerations there are again about disclosure that it is a bot in the first place and also that the conversation can be stored.
38:39
You're going to want to make sure that the place that the conversation is stored is somewhere where you can access it so that you also have the ability to delete it.
38:50
And I know I beat a lot of folks over the head with this, but when it comes to things like cookie banners, those aren't always a solve-all kind of solution.
38:59
And even when they do connect with more things on the backend than cookies and pixels and tracking technologies, you're going to want to make sure that it does
39:08
actually work when you flip the switch because I've seen enough instances, just this morning actually, of a
39:17
client who believed that their cookie banner was effective for a chatbot on their website.
39:27
Because they believed that the chatbot may have an embedded tracking technology that is taking other
39:35
information from the web page, like consumer clicks on other parts of the page.
39:40
And that just wasn't being reflected in the before and after of clicking Reject on the cookie banner.
39:46
And so really working with the technologists here, working with your web developers on understanding what actually happens on your website and doing an audit of what
39:55
tracking technologies you still need versus ones that you just have left over from the last person who set it up or from last year,
40:03
but it's not necessary anymore.
40:06
Just doing a big data cleanup of what information is collected will save you a lot of hassle in the long run when it comes to AI compliance, privacy compliance, risk management.
40:19
And I saw a question when doing this presentation yesterday on the true Data Privacy Day.
40:26
But someone asked what makes for a good professional in this space.
40:33
And I would say it's not necessarily, you know, having all your certifications or being...
40:50
It appears that we are experiencing some technical issues.
40:52
Again, I will let Anokhy know that we may have just lost her. Hopefully she answered most of your AI questions.
41:07
In the meantime, we are going to move forward.
41:13
And as a recap for the session, just a reminder that AI and data privacy are fundamentally interconnected,
41:21
that compliance is non-negotiable, and to try to follow the business best practices.