This document and the resources have been adapted from the original by James Barnett and Lisa Bird, Libraries and Learning Resources at the University of Birmingham to help researchers make informed decisions when choosing AI tools for their research. It is licensed under CC BY 4.0: creativecommons.org/licenses/by/4.0
Image by Rosy / Bad Homburg / Germany from Pixabay Accessed 19.09.2025
When choosing AI tools, it's important to assess whether they genuinely support your research needs. Consider the necessity of the tool, the quality and potential biases in its training data, and the cost of access and use.
To guide this process, we recommend starting with the Evaluative Framework for AI Tools, which helps you reflect on key issues—from relevance and effectiveness to compliance with institutional and stakeholder policies.
Before registering for any new tool, always review its Terms and Conditions (licence). This legal agreement outlines who can access the service, usage restrictions, costs, and the responsibilities and liabilities of both you and the supplier.
To support informed and responsible decision-making, please look at the two AI Tool Licensing and Review Guidance Checklists for researchers and libraries to use. These resources help you evaluate licensing terms and ensure the tool aligns with your research and ethical standards.
The Quick Review Checklist helps you rapidly assess the terms and conditions (licence) of any AI tool you're considering using. It outlines ten key points to review before registration, designed to protect you and your organisation from potentially problematic licensing terms.
An Excel version of the Quick Review Checklist is also available (Excel - 21 KB).
Provide the name of the AI tool under review.
Describe the tool’s primary purpose and key functionalities. Indicate whether it is web-based or requires installation. If installation is required, consult the IT Service Centre, as there may be security implications.
You will normally need to prompt an AI tool which may be a mix of your own words and may also include the work of others. Some tools allow a user to upload their own items, detail here any information about how they store those prompts, files or use the data from them. Some tools discard the prompts and inputs, but others will retain this data and may also use it to further train the AI model. If they are retaining the prompts or inputs why are they doing this? If they are you need to consider carefully if that is okay. You are more likely to encounter intellectual property and data protection issues if they are retaining the prompts, files or data.
Specify any IP warranties you are required to provide. For example, if you (or your users) warrant that only content with appropriate IP rights will be uploaded, and this is breached, the vendor may seek damages.
Does the tool state that you (or your users) need to be the owner of the copyright/intellectual property of any content that is loaded into it or used as an input? If yes, this would restrict what you could input to the tool to content where you are the copyright holder or have permissions from the copyright holder for this use. Take a look at the What is copyright? section on the left for information about copyright.
Does it state that you provide them with a licence to use any content that you put into the tool? If yes, what type of licence is that? What purpose do they want the licence for? Do you (or your users) have the right to give someone else a licence for the content that you are loading into the service? Normally you would only have the rights to provide a licence to material where you are the copyright owner. You would not need to be concerned about this is the material was no longer in copyright. You also need to consider data protection issues. Consider how inputting and providing a licence may impact on commercialisation of the research or the ability to publish in your (or the users) preferred places of publication e.g. by prior disclosure which could prevent a patent application.
Does the tool give any restrictions on how any outputs that are generated can be shared or used? Detail that information here. Will you be able to use the tool for the purposes that you want to use it for with these restrictions?
Confirm whether the tool complies with the current UK Public Sector Bodies Accessibility Standard (WCAG 2.2 AA). This is a legal requirement for web services provided by universities. Do not assume that users will not require accessible services, as many individuals do not disclose disabilities.
Avoid licences that require you, as the licensee, to indemnify (protect) the licensor. Such clauses may expose you or your institution to uncapped legal liabilities e.g. an uncapped financial penalty. Typically, university e-resource licences involve the licensor indemnifying the institution. Licences where you indemnify the provider of the resource can normally only be agreed to and signed by a limited number of senior ranking staff in an organisation. Contact your licensing or procurement teams before agreeing to licence terms where you indemnify the licensor.
Does the licence comply with GDPR and UK Data Protection legislation? If not, or if it permits sharing, reuse, or resale of user data without consent, seek advice from the Data Protection Officer.
Summarise your review findings. If you spot issues like accessibility non-compliance, data protection risks, or problematic indemnification terms, contact the IT Service Centre or your licensing/procurement team before proceeding. For copyright concerns, reach out to Donya Rowan, the Copyright Officer.
The AI Tools Full Licence Review (Excel - 24 KB) is aimed at supporting staff members that deal with the procurement or reviewing licences of AI tools on behalf of a group of users. This tool contains additional fields of information to be evaluated for a more in-depth review. The AI Tools Full licence Review is available as an Excel document so that you can use the separate columns as an easy way to compare the terms of a number of tools.
Though not normally covered by the licence, you may also want to ask a potential supplier of an AI tool for information on the energy consumption or carbon footprint associated with training and operating their AI models. It is good to ensure that you are selecting the most environmentally sustainable tool that meets your needs and aligns with organisational sustainability policies.
