AI is at the frontier of the workforce, and people are excited about its capabilities without pausing to consider the dangers. Right now, companies worldwide are sending personal, sensitive, and company data into AI without thinking twice about the repercussions. If your customer list gets out, your competition gains an edge, and it could create a PR disaster for your clients.

In a previous article, I wrote about the benefits of using AI this way, like uploading a contract and having AI review it. I’m not contradicting myself by saying you should never do it. I’m saying we need mechanisms in place before you do it.

Three questions people blur into one

We need to figure out what “protected” means. What does “safe” mean when you upload documents? People often blur three really big questions into one. Does the tool train on the data I give it? Who at the AI companies can see that data? Where does it live, and how long does it stay there? Those three questions define your data safety landscape.

Does the tool train on your data?

When you use a free model, like OpenAI or Claude, and don’t pay anything, that data is likely used to train their models. Here is what that means: When you ask a question or give it data, it responds. When you challenge it, it uses that to say, “Uh-oh, why did I get that wrong… why did the user challenge me?” and learns to get better and stronger. They claim they anonymize the data to keep sensitive information from getting out. But they have no contractual obligation to do that. You’re not paying anything. They have no reason to protect you.

This goes for some consumer AI as well. Even if you pay for it, there might be a checkbox you have to turn off to make sure the AI isn’t training on your data. That’s not the end of the conversation; that’s just the beginning.

Next, check your business account. If you have a business and you’re using AI, you should have a paid business account with your provider. Again, check your settings for a toggle to make sure any data training is turned off.

Once AI uses your data for training, it becomes their data, not yours. There’s no deleting. There’s no removing. You can’t ask for it back.

This ties into the other articles I’ve written. Here’s why. You can’t protect any data you can’t find. If you don’t know where it lives or who has access to it, you can’t even know what your team put in, because there’s no central repository to audit.

When I wrote about insurance questionnaires, one question is “how you protect confidential data and control access to it”. If an employee pastes confidential data into an AI model you don’t pay for, or that has training turned on, you should answer no on that form. Suddenly, your risk factor goes through the roof.

Then we have to talk about how AI can use the data inappropriately. In the first article, I wrote about how AI deleted a bunch of data out of a system we gave it access to. This is AI doing the same thing in a different pattern. It’s doing things beyond what you want it to because it has limited context or a rule set you can’t control.

Who else can see it?

As for who at the vendor can see it, that’s really unknown. We’ve seen many stories from other companies about employees accessing servers and looking at data. So far, we haven’t seen anything major from the big AI players, and they often build these so employees can’t do that. But again, it’s a trust factor. That doesn’t mean we need to stop using the tools. It means we need to understand the risk and accept the risk for what it is.

How long does it stay there?

Here’s the biggest piece. How long is your data stored on their servers? Personally, I’m not so worried about an employee at these companies looking at data. I’m more worried about a bad actor breaching an AI company and exfiltrating the data. Now all those documents, files, and chats you had are theirs. We know they use strong encryption and keep their infrastructure secure, but we’re all human, and we make mistakes. Heck, even AI makes mistakes.

So the last question is: how long do they store that data before deleting it? In technical terms, we call that a retention policy. Most of the large AI companies let you set a retention policy, but only if you pay for the enterprise version, which gets costly.

Not all data is sensitive

Here’s the flip side. Not all data is sensitive. AI drafting an email, helping you set up a job post, or answering general questions about business ideas doesn’t really need protection. If your team is using AI for those things, I wouldn’t worry about locking them out completely.

That said, being paranoid about everything is how you end up with what we call shadow AI. People use it anyway, but they hide it from you by setting up their own free accounts. Letting people use AI in your company in a controlled way is better than letting them use it on their own, uncontrolled. The goal is to distinguish between safe and unsafe, and to accept or tolerate the risk rather than getting rid of the tool altogether.

What to do this week

As always, I want to leave you with some action items.

First, find out which AI tools your team uses. Ask them. Let them know the goal is to improve it, not take it away. You’ll probably be surprised by what they’re using and what data they’re entering. In most cases, it’s not malicious; it’s because they just don’t know.

Second, make sure you’re paying for the business tier of whatever you’re going to use. Read the terms about training. Make sure it’s turned off and deployed to your team. Yes, it costs money. But the inverse costs your business.

Third, draw a really simple line. Here’s what’s allowed to go in. Here’s what’s not. This depends on what you pay for, your risk tolerance, and the model’s retention policy.

Fourth, connect any AI model you’re using to your tools directly, not by copying and pasting. Why? Because when you connect it directly, you can control what it can and can’t do. When you copy and paste, you’re just copying and pasting. Having someone copy an entire email thread that might include financial figures you don’t want to share is a bad idea. Giving AI access to that person’s email, with constraints on who it can read emails from and who it can’t, is a little better.

And lastly, think about this. If you’d be embarrassed to see your business in a headline about something you put in AI, it probably doesn’t go into a chatbot. Unless we’re talking about a specific enterprise use case where you can control the retention policy. Even then, the risk is still there, but it’s very low.

Here’s the path: set it up right, and your team can use the tools safely and securely, without having to be paranoid about security.

Remember, the security work you do isn’t supposed to be the boring stuff before you get AI working. It’s what makes AI safe to use on anything real in your company. It’s about knowing who holds the keys, what insurers require, where your data lives, and how long you retain it. If you set this up right and secure it, AI stops being a gamble. It becomes a powerful tool for your business.

In two weeks, I’ll get into automating a broken process. Most businesses have existing, broken processes they want to automate. Automation mimics you and your process, good or bad. If that sounds interesting, subscribe so we can send it to you.

If you have a tech decision in front of you, like AI, automation, security, whatever it is, and you want a straight read on it, that’s the kind of thing I like doing. You can grab some time with me here.

Why the forms got harder

Why, over the last five to ten years, have these forms become a lot harder? Because insurers have gotten burned. A lot of businesses are getting breached and filing claims that the insurers had to honor. And of course, the game of insurance is to get the clients and not pay out if they can help it. That’s exactly what they’re starting to do.

I was looking over one of the first forms I filled out. I actually got a kick out of how simple it was. They’d ask you questions like, “Have you ever been breached before? Do you have backups?” A little bit about your business to understand your risk. Are you using multi-factor authentication? What antivirus are you using? Do you have a firewall? Some of the basic stuff you’d ask a client if it was 2016. Cyber breaches weren’t as prevalent as they are today.

I compared a form from the same insurance company from 2020 to now, and the pages have gotten much bigger. The form used to be only one to two pages, and the questions were pretty simple. Now, in some cases, it’s over ten pages.

Here’s the problem with filling out that form and sending it in without really knowing the answer. If you check yes to a box (let’s say multi-factor authentication because I see that one a lot) and you get breached, the cyber insurance company will step in to help you with the legal, the forensics, and the IT side. But part of the forensic work is also determining how the breach happened. If it happened because multi-factor authentication was off, or the user who got breached had it off, they can deny the claim. And now that claim, which I’ve seen be upwards of half a million, is denied and is completely out of pocket to you. You have a choice: do I pay it and keep my business running, or do I shut down?

Not to get too deep into the cybersecurity side, but the bad actors know how much money you can pay. Before they come up with the ransom, they look through your documents. They look for bank statements and other documents to see how much you actually have.

I don’t have one specific story, because I’ve seen many claims over the years. Still, filling out the forms honestly doesn’t automatically deny you from having cyber insurance. In fact, it often doesn’t. It just puts you in a different risk category, and it does cost a little more per year because the insurance company has to take on more risk with you.

What they’re asking for now

Here are some things they’re asking for today. Some are the basics, things you’ve been asked before, and some are a little more advanced.

They’re asking whether you have a managed IT provider, and if so, they want information about what they manage for your business and what you’re paying them for.

Some are asking how many unique records you have that contain sensitive personally identifiable information, and how many are non-sensitive. Things like email addresses, phone numbers, names, which are technically personally identifiable information, but not sensitive.

They’re asking what kind of endpoint protection you use on your computers, so antivirus, as an example. They’re asking if you use endpoint detection and response, and if so, what it is. They want to know if it’s deployed to all endpoints on your network. That word “all” is really crucial. If not, they want a percentage and why it isn’t deployed everywhere.

They want to know about your perimeter security. Your firewalls (if you have them), whether you’re paying for the services, whether multi-factor is turned on to access them, whether you scan through the firewall, and whether you conduct penetration testing on your network. If you don’t know what penetration testing is, that’s okay. Take a second to search it online. It’s intensive and can get expensive, and they want to know.

Multi-factor authentication enforced on any remote access and on all company email accounts, which in many cases is just your 365 or your other email provider. Whether you simulate phishing attacks to test employees at least once a year. Whether you use a separate email filter, and what it is.

In some cases, they even ask for your Microsoft Secure Score if you use Microsoft 365. That matters because many people don’t make the necessary security changes in 365, and their score is around 25 out of 100.

If you have Active Directory, they want to know whether non-IT users have local administrator accounts and whether they have local administrator rights on their computers. They want to know whether any network monitoring detects suspicious or malicious behavior, separate from the firewall. They want to know if you have any end-of-life software, meaning software the vendor no longer updates, that could cause security vulnerabilities.

Patch management. In the past they’d ask if you did it. Now they want to know how you ensure that all critical patches are applied. They want a timeline of how quickly you’d handle things like zero-day vulnerabilities. They want to know any major changes you have planned in the next 12 months. Not ones you’ve done, the ones you’re planning.

They want to know how you protect confidential information. Things like encryption, network segmentation, access controls. They want to know how often you purge records that are no longer required. Data retention is another way to say that.

Backups. They want to know how you store your backups of critical data. Specifically, how they’re stored, not just whether you back up. They want details on the frequency, including how often you take a full backup and how often you take incremental backups. Meaning how often you take a full copy of your system, versus how often you’re just updating that copy with the newest data. They want details on how you secure those backups, and how you test them. How do you make sure that if you have to recover them, they actually work? They also want to know how many backup copies you keep and how you prevent separate copies from being impacted by an incident.

They want to know if you provide training on social engineering. And although this doesn’t sound like cybersecurity, they want to know that before you transfer funds to an account you haven’t paid before, you get authorization from the recipient through a different method than the original request. Meaning, if they send an email saying, “Hey, please update our bank account,” you call them on a number you already know and ask them to verify the bank account information verbally.

Then they’ll have a bunch of control boxes you can check to say you have these things in place. Web content filtering, for example, or an intrusion detection system. Those have been there before, but now they carry more weight because they want the name of the software for each thing you check. If you check “Yes, I’m using a DNS filter,” they want to know what product you are using to do that.

They’re not asking to be nosy, per se. What you fill out here are the things they have seen cause security breaches. They want to protect you against the things you didn’t know to do, the things you shouldn’t have been expected to know how to fix, or the things that aren’t industry standard today.

A free security audit

Look at it differently. These forms are free security audits.

Often, even if you have a provider and send it to them, they might not fill it out truthfully. But if they fill it out truthfully, it’s a good indicator of whether or not they’re providing what you need for security, or whether they can provide it and you’re not paying for it.

And if you can’t answer a question on one of these forms, it doesn’t really mean there’s a problem with the insurance company or the form. In many cases, it’s a gap in your company’s security.

What to do with it

Don’t guess. Don’t check yes because you want it to sound better to the insurance company. And certainly don’t check yes if you don’t know. Find out. If the answer is no, put no. That’s very important.

If you have someone in a technical role, take the questionnaire to them and go line by line, and have them answer those questions for you. Anything you can’t check yes to becomes something you need to work on in your business to make it more secure. And by the way, once you do that, your premium will lower.

Being honest now saves you from paying out hundreds of thousands of dollars later on a claim.

The other side

Insurance companies do have tiers, and it’s also based on the type of business you have. They may ask if you have certain things, like custom threat intelligence. Most businesses I’ve worked with don’t need custom threat intelligence, because they don’t hold classified government secrets.

So don’t feel like you have to have everything on the form. Use your judgment, and rely on technical resources to tell you what you should and shouldn’t have. Often, the forms let you explain your reasoning for checking no, and if you have a good enough reason, you’ll still be put in the lower-risk category.

This isn’t about buying everything on the list. It’s about mitigating risk, which is ultimately your job as a business owner or business leader. So when you get these forms, fill them out honestly. If there’s a gap, research it and fix it to prevent a breach in the first place.

In two weeks, I’ll get into the truth about your data and AI. Most businesses will blindly upload data to an AI model, but have no idea if they’re protected. If that sounds interesting, subscribe so we can send it to you.

If you’d rather talk about risk management and how to protect your business, that’s the kind of thing I like doing. You can grab some time with me here.

The business owner brushed it off and didn’t seem terribly worried. We had a good discussion after that, but you could tell they didn’t take these things seriously.

Months go by, and I get a phone call from that business owner. He called to tell me that this person had passed away. They were panicking because they didn’t have any of the information they should have: logins to systems, or even how to set up basic things like a new employee. Of course we offered to help, but they turned it down, and that’s where the story ends. Unfortunately, I don’t have anything more.

But it’s really, really important that I tell you that story. Because although it’s the extreme side, the ordinary version of this happens all the time. People quit. People get fired. Maybe they get sick, and they’re on sick leave. Or maybe they’re just sick of you, and they leave. The problem is that businesses find this out way too late. They find out there’s only one person who can log into a system, only one person who can make changes. And when that person leaves, they hold the keys to the kingdom.

Getting everything back can be somewhat destructive and will take weeks or months. Meaning if you needed to hire an employee and get them an email address spun up, you couldn’t for at least a couple of weeks.

Although this is framed as a technology problem, it’s really not a technical problem. It’s a who-holds-the-keys problem. Sometimes those keys are stuck in someone’s head, and whether they’re around to give them to you, or not willing to give them to you at all, is a moot point. The business owner needs to mitigate that risk altogether.

So let me give you a quick 20-minute check.

The 20-minute check

Let’s start with the really simple one. Your website domain, whatever yourcompany.com is. Who owns it, and who can log into it? Do you need two-factor authentication to access it?

Your email provider: Microsoft 365, Google, or somebody else. Who has the administrative access to that?

Then look at things like your website, accounting software, ERP or CRM, phone systems, network passwords, and server passwords. Who controls all of those?

On each account, whose email is on the recovery email or the recovery phone? In some cases, if you need to make a change, they require a second step of verification.

And the last question you should ask about every item: if this person vanished, do I have a way to get into it that doesn’t go through them?

What to fix today

You, as the owner or the person in charge of this crucial piece of the business, need access to everything. Or, if you have a contractual obligation with a tech firm not to have access to everything, they often have a sealed version of how to get in. What do I mean by that? A lot of IT people, and even IT companies, don’t want you to have admin credentials, for good reason. They don’t want people making changes to systems they don’t know about that could cause security problems or outages. But they’re happy to give you a sealed version. Once you break the seal and get in, whether through technology or other means, the IT person gets notified, or, in terms of a contractual obligation, there are clauses of liability and cost in case something breaks.

Another thing to think about is recovery contacts. Make sure the email and phone number point to a business system, not just a person. Set up a shared number through your phone service that can receive text messages, such as a Google Voice number, or even something as simple as a cell phone that sits in a drawer. As much as I don’t like that idea, it’s better than having it on someone’s personal cell phone that no one can access.

More than one person needs to know how things work. Even if it’s documented in a binder. If you want to print it because you’re old school, that’s fine. But document how everything works, and make sure more than one person knows it. The document should also include what’s where, how it connects, and how one system talks to another. That way, if something happens and you want to bring on another technical resource, you can hand them these pieces. It’s not a three-month discovery process of finding problems. It’s a one-day read, and business continues as normal.

The other thing to fix today: figure out what you’d actually tell a client if you couldn’t get into one of your systems. Or what you’d tell a new employee, or an existing one, if something happened and you couldn’t get in.

The other side

So that we’re not playing on fear, it’s not about trusting your people, because most of the time nothing happens. You can trust someone completely and still not want them to be the only key. If you have a spouse, they have the key to the house too.

It’s not about the business being afraid of someone doing something wrong. It’s about the business being one major event away from being locked out of everything it has.

Doing this will take you twenty minutes. But even if it took you an hour, that hour might save you weeks or months down the road.

In two weeks, I’ll get into the cyber insurance landscape. The questionnaires you get are more important than you think, and they’re getting tougher. If that sounds interesting, subscribe so we can send it to you.

If you’d rather talk about risk management and how to protect your business, that’s the kind of thing I like doing. You can grab some time with me here.

As we go along, more and more features keep getting added to these tools, but their full potential is still unknown, including the AI ones. I think people are starting to ignore these assistants, since early uses have shown they’re very rudimentary and not extremely useful.

So as I’ve gone along in this journey of figuring out what tools work and what tools don’t, I have made some observations, and I’m going to give you the rundown: what things actually matter, where it falls short, and what’s going to cost you long term.

What actually matters

My definition of useful here is that anything that helps you save time or money in your business. That’s the biggest indicator of whether a tool should be on your radar or not. That looks different for everyone, but there are a couple of ways we use it today that can help us.

The first one is drafting things. Proposals, emails, job posts, policies we have to write, graphic design content, and even figuring out how to word a specific message for a specific audience. It takes you a good portion of the way, but it doesn’t get you to the finish line unless you want to accept whatever AI delivers to you. This isn’t a good idea. A lot of people do this, and because of that, the digital world is filled with what’s called AI slop.

Another helpful thing is pulling answers from your own data, if you can feed it in. For instance, we pulled financial data because we could get our AI models to connect to our financial systems. We asked very simple questions, like which companies are costing us more money based on the hours we put in. We asked which vendors have been slowly increasing costs on us without us realizing it. What’s our average cost of goods percentage year over year? Things like this help us figure out where we need to make adjustments, without spending a day on spreadsheets or making data-entry mistakes.

If you want to try this yourself, you can type things like:

“I run a [type of business]. Ask me ten questions about how my money is spent in my business, and tell me where I’m losing money, year over year or even month over month.”

“Here’s a vendor quote. Give me in plain English what this quote is, flag anything that doesn’t look right, or terms that don’t stand well with my organization.”

For instance, if the vendor is locking you into a five-year auto-renewing term, that’s something you’d want it to flag.

But just as a note. Before you put any data into an AI model, make sure that it’s private, and that they don’t train it on what you give it. Usually that means paying for AI. Paying for an account with one of the big providers and reviewing the contract terms helps ensure your data isn’t being sold or used for training.

Where it falls short

The biggest thing I’ve found is that it does a couple of things really poorly.

First, it’s super confident in its answers, which can give you a false sense of truth. Whenever it gives you answers based on data, I’ve caught it making some of the data up. I have gotten flat-out fake data when, in reality, that data didn’t even exist. Consider challenging the model if something feels off, and give it data to prove you know what you’re talking about. Otherwise, it’ll take advantage of your blanket trust of the model.

The other thing I’ve found is that AI can actually make changes if it doesn’t understand what you want it to do. I’ll give you an example.

We had AI make some changes to one of our systems, but it didn’t fully understand the format it needed to push. So instead of updating a couple of line items, it wiped a big chunk of data for a sales order process. It didn’t give us any confirmation. Should I do this? No warning. It was very confident it would make the appropriate push to the system. When we found out it deleted everything, it made us realize how cautious we have to be about letting AI make changes.

Now, a little side note. We had everything backed up, and we always have a back-out plan before we let it touch anything. Also, the data it touched was internal data that didn’t really affect the business, and we instantly rolled it back, which was part of the plan anyway.

The ultimate problem is that the AI didn’t know what it didn’t know, and it didn’t stop and ask, Hey, are you sure? It had the data, it had the parameters, and it thought it had what it needed, but it never tried to dig deeper or ask us to do a little more work to prove it was good. Keep in mind, some of that was also because we didn’t prompt it properly. We could have said: “Before you do anything, make sure you’re certain about the change you are going to make, and verify it with known documentation; otherwise stop and present your facts and allow me to make a final decision before pushing a change.” Ultimately, if we’re going to trust AI as a business owner, not as a technical resource, we want it to do the task when we ask it to. We don’t want to quadruple-verify that it knows what it’s doing; otherwise, we could do the task ourselves.

The hidden costs

Like anything, nothing is free, and using these built-in AI models has hidden costs.

The first one is learning. Learning what it can do, how to prompt it, what its strengths and weaknesses are. That costs you time, and it makes the subscription the cheap part of the whole deal. You have to learn the tool, set it up in your workflow, and clean up all the big and small mistakes it makes. Just in time for another tool to come along that you want to try. Not to discourage you from attempting any of this, because if you can get it working, you’re going to save yourself a lot of time. It’s just a fact of any software.

That leads to the next point: something you pick today might not be the best thing in six months, or even three months from now. Companies keep updating models. They keep changing the price. They keep killing features. They keep adding modifications and safety nets to their models. You may build something that really works for you today, and in three months either something better comes up, or a really powerful feature no longer works because of some regulation or safety net they put on. That has nothing to do with you, but it does affect you.

Make sure you understand your processes. Meaning, if AI builds something for you that runs some automation, make sure you know at the very least the bare bones of it. It does this, then it does that. That way, if you move to a new model, you can rework it with knowledge instead of being lost on how AI built something, and you have a base to give the new model. “This is how this system worked,” or at least how the flow was supposed to go, instead of re-explaining your entire business use case. Documentation is really important. You can even have AI build the documentation. But I’d encourage you to make sure you know what the system is doing. In the end, it just means keeping your processes understandable, which is a good business practice overall.

Lastly, don’t lean on one tool. Like any other company, if you lean on one thing and that one thing disappears, you lose your entire business workflow. That’s a huge problem, especially if you start relying on it in place of people working. Another possibility is that you get into a dispute with the company that created the model, and you can’t use it anymore. Whatever it may be, you and I have both been through business problems like that before, and relying on one model and one system, at least right now with how new everything is, is probably a bad idea.

That’s the first one. In two weeks, I’ll bring you the 20-minute check that shows you who could lock you out of your own business, and what to fix today. If that sounds interesting, subscribe so we can send it to you.

If you’d rather talk about AI and how it fits in your business, that’s the kind of thing I like doing. You can grab some time with me here.