The Cost of Free AI Tools: three questions to ask
Free AI tools cost nothing to sign up for, which makes them an easy answer to the question "do we need to pay for AI?" If staff can already use ChatGPT, Gemini, Copilot or Claude on a personal account, why buy licences?
It's a fair question, and for some businesses starting small might be the right call. But the price tag isn't the only cost. When a tool is free, it's worth asking what the provider gets in return, and what your business is handing over to get it.
In my AI governance workshops, I walk attendees through a series of everyday scenarios. For each one, I ask the same three questions:
Where does that information go?
Who owns it afterwards?
Could it be exposed later?
They're simple questions, but they're often enough to change how a room thinks about the problem. If free AI tools are part of your plan, these are the questions to ask first.
A building somewhere in Auckland, can’t remember where
Two ways free AI ends up in a business
By decision. Some businesses choose it. Rather than sign up for business AI tools, they let staff use personal, free accounts. The saving is real and easy to see. What the business understands about the trade is often less clear.
By default. Others never made a decision at all. No one has provided an approved tool, so staff have found their own. They're trying to get work done faster, not cause trouble. Here the business may not know it's exposed.
The three questions apply either way. The difference is whether the business has taken on the risk knowingly, or had it taken on for them. And if the answer to "is anyone here using free AI tools?" is "not that we know of", that's often a statement about visibility rather than behaviour.
Where does that information go?
When someone pastes text into a free AI tool or uploads a file, it leaves your systems and lands on the provider's. What happens next depends on the provider's terms and the account's settings, and both often differ between free personal plans and business plans.
Things worth knowing:
Is the content kept, and for how long?
Is it used to train or improve the provider's models, and is that switched on by default?
Where in the world is it processed and stored?
Think of a few ordinary scenarios. Someone uploads a customer spreadsheet to tidy up the formatting. Someone pastes in an email thread from an employment dispute to help draft a reply. Someone drops in a supplier contract and asks for a summary. Each takes thirty seconds and saves an hour. Each also sends business information somewhere the business may not be able to trace.
The settings that control all this sit on the individual's personal account. The business can't see them, can't set them, and can't check whether they've changed.
Who owns it afterwards?
Ownership has several layers, and free tools often blur them.
The input. What rights does the provider take over what you submit? Some terms grant the provider a licence to use content to improve its services. That isn't the same as owning it, but it isn't nothing either.
The output. Who owns what the tool produces? Terms differ between providers, and the legal position on AI-generated content is still developing in many places. If the output ends up in client deliverables, code or published material, it's worth knowing where you stand.
The account. This one is easy to miss. A personal account belongs to the person, not the business. The conversation history, uploaded files and any custom setup all live in it. When that person leaves, the business has no claim on the account and no way to retrieve or delete what's inside.
Your obligations to others. If a customer shared information with you under a confidentiality clause or a data agreement, does that agreement allow it to be pasted into an AI tool?
Could it be exposed later?
Exposure isn't only about someone breaking in. It can happen in quieter ways:
Weak protection on the account. Chat histories sit behind whatever the individual chose: perhaps a reused password, perhaps no multi-factor authentication, on a personal device.
Sharing features. Some tools let users share a conversation by link. A link that gets forwarded can reach people it was never meant for.
Surfacing inside the tool later. What goes in doesn't necessarily stay put. Depending on the tool, it can come back through chat history, through features that remember details across conversations, or through a search of past chats. Anyone who can get into that account, whether a colleague on a shared login or someone who picks up the device later, may see what was entered months earlier.
Former staff. When someone leaves, the personal account leaves with them. They can keep signing in and reading their full chat history, including anything they entered about your customers, suppliers or business. The business can't revoke that access, because it never controlled the account.
Then there's the follow-up question: if something was exposed, would you be able to find out? A free personal account gives the business no audit trail and no admin logs, so there's no record of what was shared or when.
Making the decision with open eyes
The point of the three questions isn't to conclude that free tools are always the wrong choice. It's to make the choice knowing what it involves. Asked honestly, they lead to some practical next steps:
Find out what's already happening. Ask staff, without blame, which AI tools they use and what for. If you've never asked, the answer may differ from what you'd assume.
Decide what never goes into a free tool. Customer personal information, financial data, contracts, HR matters and anything covered by a confidentiality agreement are a good start. A short list is often easier to follow than a long policy.
Compare the real costs. Business plans often come with admin controls, clearer terms on data use and accounts the business owns. Is the licence fee more or less than the cost of one incident? That's the comparison worth making.
Give people an approved option. If staff reach for free tools because nothing else exists, is a sanctioned tool with clear guidance a better answer than a ban?
The saving from a free tool is easy to count. The three questions help quantify the risks.
We use the AI Contribution Scale defined by Blair Enns to disclose the use of AI in our written content. This article is rated as: AI-4: AI Drafted. The content was drafted by Claude from our own content and ideas, we then refined the output.