Over the past few years, AI has been rapidly gaining momentum and becoming a popular assistant across various industries. Businesses are increasingly using AI as a legal assistant for minor — or not so minor — tasks. From a business perspective, the obvious advantages of using AI as a mini-lawyer are clear: it is fast, always available; it is convenient, handling any request “well enough”; and, one of the most important factors, it is cheap.

We are increasingly witnessing businesses turning to artificial intelligence with requests to generate Terms of Use or Terms of Service, only to then encounter a range of problems that could have been avoided with quality Terms.

To the average user, Terms may often look like just technical text, but for business owners and lawyers they are not merely text — they are a carefully designed legal mechanism, and this is precisely where AI fails.

In this article we will break down the biggest AI mistakes in writing Terms & Conditions, as well as where it can be safely used.

The Core Problem: AI Does Not Understand Your Business

When a request comes in to write a legal document, AI works with generalized templates rather than a specific case.

In order to draft a quality mechanism in the Terms, a lawyer needs a thorough understanding of the product, its functionality, and other characteristics. Without such understanding, the provisions will not fully cover the service or product and will not effectively protect the business from risks or complaints. Artificial intelligence, in turn, cannot fully “study” the service for which Terms of Use need to be created.

There are countless types of services and products — SaaS platforms, marketplaces, dating apps, and others — but AI-generated Terms will look the same in all of the above cases, since AI will not focus on the specifics of each individual situation.

Let us take a dating app as an example. The first version is an AI-generated section; the second is written by a lawyer:

AI-generated:

App Functionality The App provides users with access to various features and functionalities designed to enhance the overall user experience. Users may create an account, interact with other users, access content, and use additional tools made available through the App from time to time. Certain features may be subject to limitations, conditions, or availability depending on the user’s location, account status, or other factors determined by the Company. The Company does not guarantee that any specific functionality will be available at all times or that the Application will meet the expectations of any particular user.

In this case, the formulations are excessively broad — such as “various features” and “additional tools” — which provide no specificity or understanding of the product’s nature, since AI physically cannot download and explore the application’s features.

Written by a lawyer:

App Functionality The App allows Users to access the features for social interaction, matchmaking, and online dating. a. Profile. After signing in to your Account, you will be prompted to create and customise a user profile before accessing the core features of the App. Your profile may contain various details, including (without limitation) your name, a brief description, gender, location, and any other information the App may request or allow you to add. The App also lets you upload personal photos and other media and, where available, display certain content from linked third-party services (e.g., social media accounts). Please note that your profile information and any content you choose to share may be visible to other Users. b. Search. The App may suggest and display potential matches using automated tools and algorithms based on the information provided by you and other Users in your profiles. Depending on the features available at the time, Users may also browse, search for, and view other Users’ profiles. c. Matches and Interactions. Where a match occurs — Users may interact in different ways, including (without limitation) exchanging messages, viewing or sharing photos, videos, and other media, and accessing additional profile information made available by other Users.

In this example the functionality is described in detail, and it is immediately clear that this is a dating app.

Such formulations are precise and do not mislead the user — meaning that by accepting them, the user understands what they are actually agreeing to.

AI Cannot Work with Principles and Evaluative Concepts

Often, when a user requests the broadest provisions permissible under applicable law, AI generates something along the lines of:

“We may terminate your account at any time for any reason” “We may retain content and data as long as necessary for business purposes”

Again, the formulations are excessively broad and may conflict with the law.

This also illustrates that AI cannot work with concepts such as proportionality, balance of interests, reasonableness, and similar principles — concepts that are familiar and intuitive to lawyers but are almost beyond the reach of AI.

Another example: laws frequently contain an evaluative concept such as a “reasonable timeframe.” Such a timeframe is determined by reference to established practice on a particular matter, but AI does not know these details and will therefore most likely suggest a timeframe that sounds “reasonable” but in practice is not. As a result, the company may face the problem of its documents being non-compliant with the law.

AI Confuses Laws and Creates False Confidence in Their Correct Application

This is one of the most obvious problems: a user asks a legal question and requests it to be backed by specific legislation, and receives an incorrect answer with references to incorrect or simply invented laws and articles.

Let us take the same Terms for a dating app as an example.

In the “Account Termination” section, AI generated provisions that read as follows:

Users may delete their account at any time by submitting a deletion request through the App. Upon such request, the Company will permanently delete the account and all associated data within 60 days, according to the Digital Content Deletion Directive (“DCDD”) During this period, the Company may retain and continue to use the User’s data for business, analytical, security, or operational purposes. The User acknowledges that account deletion does not immediately result in the removal of all data from the Company’s systems.

Looking at these provisions from the perspective of EU law, the first thing to note is that AI paid partial attention to EU consumer law but completely disregarded the provisions of the GDPR, while at the same time referencing a non-existent law.

Here is the explanation:

The 60-day timeframe for deleting the account and all data is not appropriate, as it was proposed without any legal logic or legal basis — the law referenced by AI does not exist.

Furthermore, the provision does not distinguish between the types of data being discussed — that is, AI uses the word “data” as a catch-all term covering both personal data regulated by the GDPR and content regulated by EU consumer law. The deletion timeframes for such data will also differ, and the best approach is to align the content deletion timeframes under EU consumer law with the personal data deletion timeframes under the GDPR.

In addition, the EU consumer law requirements referenced by AI are not fully complied with: pursuant to Article 16(5) of the Digital Content Directive (“DCD”) and Article 13(8) of the Consumer Rights Directive (“CRD”), following a request to delete an account, the trader must make such an account inaccessible and invisible online — immediately.

That is, even when AI does reference the correct legislation, this is no guarantee that it will be correctly applied in the Terms of Use.

As a result, the business is confident that all Terms provisions clearly comply with the law and protect the product — when in reality this is not the case at all.

Where Can AI Actually Be Useful When It Comes to Legal Documents?

AI certainly has its advantages, and it would be wrong to discredit it entirely.

So where can AI be safely used when working with legal documents?

AI can help with interpreting and explaining Terms & Conditions provisions to your team — it handles explanations based on already completed documents fairly well. That said, AI may not explain everything correctly, so it is worth verifying the information or seeking an explanation from a lawyer.

AI can reduce the time needed to understand the provisions of counterparties’ or potential partners’ Terms. This will of course only be a surface-level analysis, but if there are glaring violations in the provisions AI will flag them — though if the gaps are less obvious and better concealed, it is worth consulting a lawyer.

If you do decide to use AI to generate Terms, it is worth bearing in mind that AI can only produce a quality document structure — its content will require the attention of professional lawyers in order to ensure protection and compliance with applicable law.

Conclusions

AI-generated Terms of Use often look convincing, but in practice contain inaccuracies, may reference non-existent laws, and do not account for the specifics of a particular product. Such mistakes can lead not only to a loss of legal protection, but also to real risks — from consumer complaints to regulatory sanctions. AI can therefore be a useful tool at the drafting stage, but the final version of Terms must always undergo professional legal review.

The Legal IT Group team is always ready to create quality Terms of Use for your business.

Сообщение AI-Generated Terms of Use: What Can Go Wrong? Breaking Down AI Mistakes появились сначала на Legal IT group.