AI FAQs

GenAI (Generative AI) creates new content, such as text, images, audio, video or code, in response to an instruction or question. It generates outputs by identifying patterns in the data on which it was trained. Its answers can appear convincing but may be inaccurate, incomplete or biased, so you must check the output before using it.

Generative AI can help with:

• drafting documents
• summarising information
• generating ideas
• creating presentations
• analysing information
• writing code

AI does not understand information in the same way people do. It generates responses based on patterns in data and may produce inaccurate, incomplete or misleading outputs. See What are the limitations of using GenAI tools & what should we be careful about?  

For a list of AI tools we are using see What approved AI tools do we have available internally and how to access/download them?

Generative AI can produce content that looks convincing but may be incorrect, outdated, oversimplified, or incomplete. Use the Check, Challenge, Critique approach to safeguard quality and compliance: 

1. Check 

• Verify information with other authoritative sources, policies, and compliance requirements before using or sharing. 
• Check for fabricated sources: AI may create citations or references that look real but are fictitious. 

2. Challenge 

• Question assumptions, logic, and potential bias. 
• AI reflects the biases in its training data and may amplify them, so ask: 
  • "Could this disadvantage certain groups?"
  • "Are important perspectives missing?" 

3. Critique 

• Assess tone, clarity, and appropriateness. 
• Consider lack of transparency: You cannot see which sources were used or how the output was generated. 
• The tool can help (but it's your role to do final checks), ask:
  • "What evidence supports this answer? "Provide sources for any statistics, research or factual claims." 
  • "Could this response disadvantage or overlook any groups or perspectives?" 
  • "What are the risks or unintended consequences of this approach?"
  • "What is missing from this response?"
  • "Where is the reasoning weak or unsupported?" 

Be mindful of privacy risks: Public tools may store inputs for future training. (See, What AI tool can I get started with? on using the built in Copilot to reduce these risks.) 

✅ DO: 

• Act ethically: Be transparent when AI helps you create content—acknowledge its use. 
• Maintain your expertise - Use AI as a tool, not a replacement for professional judgment. 
• Write more effective prompts (instruction to AI) - Clear, precise prompts lead to better results – learn how to write effective prompts in our Generative AI Fundamentals e-learning.  
• Take ownership – You are accountable for everything you send; AI outputs need human judgement and critical thinking. 

❌ DON'T: 

• Input sensitive data Never enter special category personal data such as health information, biometric or genetic data, or any criminal offence data into any tool unless you have cleared this with the information “data owner” (either a person, or a department that has authority over a dataset).
• Use unmanaged AI tools for personal data - Do not enter any identifiable personal data into AI tools that are not approved and managed by Queen Mary. Check the AI software list on what data you can share with what tool.  
• Share confidential research Never input unpublished findings or grant applications under review. 
• Forget context - AI doesn't understand Queen Mary policies, legal requirements, or local nuances.  
• Ignore copyright - AI may generate content similar or the same as to copyrighted works - check before sharing. 
• Rush outputs - Always review and validate AI-generated content before use. This [PDF resource] can help you consider what to check. 

Entering data, especially sensitive data, into AI tools may not be secure. These services often store user inputs and may reuse them to improve their models and generate predictions, content, recommendations, output, or decisions. 

Any information shared could be retained and so may not be handled as per our compliance requirements regarding data protection legislation nor our institutional policies. This creates serious risks when handling student or staff data, or any other information covered by GDPR, unpublished research, confidential documents, grant applications.

Please see the AI software list to see what data you can share with which tool.

Put simply, you should not put Queen Mary information (including images) into publicly available tools not approved by Queen Mary for further information. See What AI tool can I get started with? and Can I use ChatGPT, Claude, Gemini or similar? 

AI tools rely on large data centres that use a lot of electricity and water and most of this energy still comes from non-renewable sources. Training and running AI systems can produce significant carbon emissions and while a single query may seem small, the combined effect of millions of queries adds up quickly. 

What we can do to help:

• Be conscious of usage: Use GenAI tools thoughtfully, balancing their benefits with the need to reduce unnecessary power consumption and the potential environmental impact.
• Prompt efficiency: Craft clear, precise prompts to reduce unnecessary iterations and save time and energy. Learn effective techniques in our Generative AI Fundamentals e-learning course. Save successful prompts and share with others.
• Reuse and refine: Instead of starting from scratch, build on previous outputs to minimise processing.
• Be thoughtful about creating AI images, audio or video – these usually have a higher environmental impact than text generation, so consider official Queen Mary images wherever possible. Any images should adhere to Queen Mary’s brand guidelines and comply with copyright and permission requirements.  

Queen Mary is committed to advancing a sustainable and equitable future and to mitigating the environmental impact, arising from the use of AI through:

• Sustainable infrastructure: Queen Mary is investing in energy-efficient data centres and innovative projects, such as using waste heat from AI-powered data centres to heat campus buildings, directly supporting our net zero carbon targets.
• Responsible procurement: All new enterprise AI solutions are evaluated for their environmental impact as part of our procurement and approval processes, in line with the Queen Mary’s sustainability policies.

Further reading:

• Queen Mary Sustainability Policy
• Centre for Excellence in AI in Education - Teaching with Generative AI: Balancing Innovation and Environmental Responsibility 
   
  
  

At Queen Mary, we have the built-in version of Copilot. It is a free AI assistant available to all staff through our Microsoft 365 subscription. It offers greater protection of our data than other free tools.

To access Copilot:

• Go to m365.cloud.microsoft/chat
• Log in with your Queen Mary account.
• Confirm that you are signed in to the Queen Mary version of Copilot—look for the green shield icon in the top-right corner (sometimes you can find it in bottom left) of your screen. Hovering over the green shield displays the enterprise data protection confirmation.

The green shield icon confirms you’re using the Queen Mary secure version of Copilot.

When you use Copilot while signed in with your Queen Mary account, enterprise data protection applies. Your prompts and responses are protected under Queen Mary’s Microsoft agreement and are not used to train the underlying foundation models. 

You can also access the built-in Copilot via Windows 11:

• Click the Copilot icon in the Windows taskbar (or write Copilot in the search box at the bottom of your screen and you will see the icon)
• You may see 2 similar apps, Microsoft 365 Copilot and Copilot.

• It doesn’t matter which one you pick – however you MUST sign in with your QM account

If you also have Enterprise Co-pilot you will have a work/personal toggle button. Keep this on "Work" mode personal mode does not include the correct data protection and should not be used with any Queen Mary data.

Always check for the green shield to confirm you are logged in to your QM Account. See our short guide on the built-in Copilot [PDF 71KB] for help and see What AI training and development courses are currently available for QM staff and how to access them?  for training available.

 

What approved AI tools do we have available internally and how to access/ download them?

This AI software list provides tools that have been approved, not necessarily supported, for use at Queen Mary and what kind of data you can share with them.

This list also provides you with links to download the tools or link to the software request catalogue. Software available through the catalogue will take around 10 days to be procured and made available on your supported device (where applicable).

If you have a supported device, you already have access to built in Copilot - see What AI tool can I get started with? for how to log in. Additionally, you have access to other free tools. See Can I use ChatGPT, Claude, Gemini or similar? for further information and how to protect data.

Never input the following data of identifiable individuals into any tools until you have cleared this with the information “data owner” (a senior individual responsible for a logical grouping of data):

• Health information
• Biometric or genetic data
• Any other special category personal data
• Criminal offence data

Why?

These categories are high risk requiring enhanced safeguards and formal review on a case-by-case basis, usually by completing a DPIA (Data Protection Impact Assessment).
 

Possibly, please refer to the AI software list to understand which free/public tools are approved for use at Queen Mary. Typically, they are fine to use for idea generation, drafting non-sensitive text, or learning about a topic.

Do not add any Queen Mary information unless it is already public information. In particular confidential research, unpublished findings, grant application, staff/student personal information, Queen Mary strategies, financial or commercial information. For these you must use the built in Copilot (check you are you signed in,  see What AI tool can I get started with? to find out how) or another tool approved for use with QM data on the AI software list. 

Also do not share imagery or designed assets into public tools to generate feedback or create new versions as this will unintentionally devalue the creative work.  

Why the distinction between built in Copilot and other publicly available tools? 

Public Gen AI typically:

• Store your conversation history on external servers
• May use your inputs for model training (depending on settings)
• Aren't covered by institutional data processing agreements

Good practice:

• Before inputting text, remove all identifiable information
• Check outputs before sharing - AI frequently makes errors
• Opt out of training the models where possible, through tool settings

 

At Queen Mary, there are typically two main versions of Copilot available. One is the built-in version, the other is Enterprise Copilot 365 which has a cost, please see AI software list.

Both Built-in Copilot Chat and Enterprise Copilot version can have 365 in the name so it can be confusing.

If you see Copilot features inside Microsoft Office apps, you are likely to have Enterprise Copilot.

You can also check on the main chat website - in the bottom left you will explicitly see the license: 

If you only have access to the web chat or do not see the “You have premium Copilot features enabled” message, you have Built in Copilot Chat (free version).

A summary of the key differences in the capability of each version follows:   Built in Copilot chat Enterprise Copilot for Office 365 Log in ALWAYS use your Queen Mary IT Account (Do not log in with personal Microsoft accounts) Log in with your QM IT Account to use the licensed version Cost Free ~£24/per user/per month Data protection Covered by Microsoft Enterprise data agreement when logged in with your QM IT Account (look for the green shield in the top right) Covered by Microsoft Enterprise data agreement when logged in with your QM IT Account Data Access Files that you upload The Web Documents in the current context of the chat Work mode: Access to all your files in OneDrive and SharePoint that you have access to as per sensitivity labelling It will use this data to ground and provide any context to questions you ask. Calendar, Emails, Teams chats, Transcripts etc Web data and Bing search. Note: When using highly sensitive data – turn off Web Search in work mode   Web mode: Access to web data Bing search. Files that you explicitly upload or reference Note: when using web mode do not include sensitive information in the prompt Availability Chat only within the app and website Integrated chats from Microsoft Teams for improved meeting and chat access, Outlook for improved Email drafting and review, Word to quickly add to documents, Excel to assist with data cleanup and functions. Chat available as per the free version Main uses Drafting, reviewing and summarising in the chat windows Personal Assistant Drafting, reviewing and summarising Searching and creating files Free access to Microsoft image creating Research and deep diving into documentation Access to create personal AI Agents (limited availability at QM currently)

Before entering any Queen Mary information into Copilot, check that: 

• your account/profile icon shows your Queen Mary email address, not a personal Microsoft, Outlook, Hotmail or other account 
• you can see the green shield icon, which indicates the work-protected Copilot experience 
• If you have the Microsoft 365 Premium Copilot licence check
  • the toggle is set to Work, not Web or Personal
  • Work IQ is enabled when working with Queen Mary information. It allows Copilot to use the Microsoft 365 information you already have permission to access, while applying Queen Mary’s existing security and compliance controls. 

 

Staff may experiment with free AI tools, for their own learning, only if the tool is used in a way that protects data privacy namely, no Queen Mary, personal or sensitive personal data is entered. 

If you intend to use Queen Mary data (such as staff, student, research, or confidential information) with a free tool, you must submit the tool for approval via the Ideas portal – Software Request before you start using the tool in line with the IS07 Third Party  Policy.

Why the restriction? Free tools rarely offer enterprise-level privacy, contractual guarantees, or data control. While they are safe for personal exploration, they are not safe for handling Queen Mary or personal data without formal approval.

If it is a “paid-for” tool please see the answer to A tool I want to use is not on the approved list so how do I introduce a new AI tool to Queen Mary?

All paid for tools go through a trial. All trials, even on a small scale, must be submitted for approval this ensures that any risks to Queen Mary’s data or systems are properly assessed and managed before use.

What you can do without formal approval:

• Make full use of approved tools for your work.
• Read vendor documentation and watch demonstrations.
• Discuss ideas and possibilities with colleagues
   

This approach helps protect both you and Queen Mary from unintended risks, and ensures that all technology use remains safe, compliant, and effective.

To suggest a new tool, follow this process: A. Check if it's already under review: Visit the “AI Tools Under Review” section on  AIIT (AI Innovation & Transformation) Hub If the requested software is already being trialled the Ideas Forum team will endeavour to put you in contact with the sanctioned trial owner and you may be able to participate. If you would like to be part of a trial please submit your request via the Ideas portal – Software Request. B. If not listed, submit a new request in the Ideas portal – Software Request (in line with the IS07 Third Party Policy), prepare your proposal -  include: Tool name and description Use case - what problem will it solve? Who would benefit? Why can't existing approved tools meet the need? Cost (if any) What Queen Mary information would you input into it? C. Evaluation - Your proposal will be assessed on: Strategic fit Security Data protection Cost-benefit Scalability D. Decision timeline - 1-2 weeks for most tools. Depending on complexity and business criticality. After you submit your request, we will provide an estimate of the timeline.

The system and its new AI feature will need to be reviewed for data protection, security and strategic fit.

If you would like to use the new feature following experimentation, please submit a request via the Ideas portal – Software Request before using the new AI feature in a production environment.

This process helps safeguard Queen Mary’s data, ensures responsible adoption of AI and supports coordinated enterprise-wide AAA strategy (see question 16 for details of AAA programme).

Depending on the cost of the new feature, you will either need to submit your request via the Ideas portal – Software Request or follow the formal procurement process in line with the IS07 Third Party Policy . All IT purchases including software extensions and upgrades must comply with Queen Mary’s IT purchasing principles and procurement cost requirements.

 Why is this necessary?

Paid extensions or upgrades need to be assessed for cost effectiveness and must demonstrate a return on investment. They may also introduce new risks or requirements that have not been assessed. The product and its new AI feature must undergo institutional AI checks, and procurement must be coordinated by ITS and the Procurement Team.

 What do I need to do?

Talk to your Faculty Relationship Manager, or if you are within Professional Services, submit a request via the Ideas portal – Software Request.

 This ensures the new feature is reviewed for data protection, security, compliance and strategic alignment with Queen Mary’s enterprise AI strategy.

 Cost thresholds apply: Depending on the value of the purchase, additional approvals and competitive tendering may be required. For full details, see the official IT purchasing principles and process and Procurement website.

Certain AI tools are prohibited due to legal, ethical, contractual or security risks. You will find these on this list labelled Not approved

How are these decisions made?

The decision is based on a formal governance process led by a multidisciplinary group including Enterprise Architecture, Information Security, and Data Governance experts. This group evaluates each tool against Queen Mary’s Policies, Principles, Strategic alignment, GDPR compliance, and contractual obligations. High-risk proposals are escalated to senior oversight for review, and all decisions are documented for transparency.

Why: 

• These tools record, store, and process meeting audio on external servers
• Data may be used to improve their models
• High risk of exposing confidential, personal, or research information
• Significant risk of data leaks

Does this discourage innovation?

Not at all. We actively encourage innovation through approved tools, vendor demos and Communities of Practice. If you have an idea or want to trial a new tool, you can submit it via the Ideas portal (in line with the IS07 Third Party Policy). This process ensures that innovation happens safely and responsibly, without exposing Queen Mary’s data to unnecessary risk.

Use alternative tools on the Queen Mary approved list or go to A tool I want to use is not on the approved list so how do I introduce a new AI tool to Queen Mary?

This could expose Queen Mary to increased risks such as data leaks. Some AI tools can store students’ personal data, staff records or intellectual property, research or sensitive organisational data outside of Queen Mary, beyond our control.

For instance, when an academic chooses to upload unpublished research onto an AI tool that isn’t approved for Queen Mary data, that tool can share that academic research with third parties or train a model with this newly found data. These AI tools may also breach research contracts or licensing terms, or even the law in relation to data processing and data storage. Finally, these tools may contain malware or spyware that can harvest Queen Mary’s data for malicious purposes.

Please refer to AI software list to see what tools are approved for what type of data.

We’ve curated all AI learning resources, guidance, and training on CPD training here. This includes introductory AI learning (including prompt writing guidance), Microsoft Copilot guidance, training focused for educators and researchers.   

1. How can I get involved in AI at QM?

There are several groups you can join, each focused on different aspects of AI and innovation at Queen Mary.

AI in Research 

Digital Environment Research Institute (DERI)
DERI leads cutting-edge AI and digital research as part of Strategy 2030.

• Collaborate on interdisciplinary projects tackling healthcare, sustainability, and AI ethics.
• Explore flagship initiatives like Living with Machines (with The Alan Turing Institute and British Library).
  Join here: https://www.qmul.ac.uk/deri/discover-deri/join-deri/

AI in Education 

Centre of Excellence for AI in Education (Queen Mary Academy)
A developing hub for promoting AI in teaching, pedagogic scholarship, and training. 

• Shape future directions for AI-enhanced learning.
• Access resources and CPD opportunities.
  Join here: General | QMA AI in Learning Community of Practice | Microsoft Teams

Technology Enhanced Community of Practice – AI Channel

Focused on practical AI applications in teaching and learning. 

• Share best practices and success stories in education.
  Join here: Artificial Intelligence | Technology Enhanced Community of Practice | Microsoft Teams

AI for Day-to-Day Productivity 

Ideas Forum:

• Submit formal proposals for new tools or process automations via the Ideas Forum

 

QMA have developed a guide, an online course and communities of practice. Find out more at: Staff Guide to Generative AI - Queen Mary Academy

Start with the Ideas Forum. This is the formal route for introducing tools or initiatives not already approved. An outcome is normally provided within 7-10 days. The Ideas Forum Group meets once a week on a Thursday afternoon to discuss Ideas, and we will invite you to talk through your proposal.

If it’s a tool you want to introduce you can check on the AIIT (AI Innovation & Transformation) Innovation Hub on SharePoint to see if your idea or tool is already under review.

We’re moving towards using automation and artificial intelligence (AI) in a way that is joined up and consistent across Queen Mary. Rather than building lots of separate systems, we’ll focus on a small number of well-supported platforms that work well together. This approach helps us keep things consistent where we can, simple and secure.

Sometimes there will be special cases, such as unique academic or research needs where a different solution is necessary. These exceptions will be carefully assessed to ensure they remain sustainable and secure.

AI and automation initiatives will be prioritised based on impact and return on investment. Projects that deliver the greatest value or support organisational change will come first. This ensures our resources are directed toward initiatives that make a measurable difference. 

By following these principles, we can deliver a consistent, secure and future-ready approach to automation and AI that supports the diverse needs of our academic community while allowing flexibility for justified niche requirements.

Current Initiatives:

• The AI initiatives are part of the AAA (Applications, AI and Automation) programme. This programme aims to embed artificial intelligence, automation and modern enterprise applications at the heart of our operations, education and research through: 
  • driving digital transformation 
  • enhancing efficiency and service quality 
  • fostering innovation through centres of excellence 
  • strengthening data governance and integration  
  • ensuring the ethical and responsible use of AI. 
• The anticipated outcomes include streamlined processes, improved experiences for staff and students, more agile and data-driven decision-making, a culture of innovation and demonstrable value through cost savings and new opportunities enabled by technology, all underpinned by a five-year strategic roadmap for AI and automation.  

• Institutional AI, QMOne: QMOne is an AI platform developed, maintained and continuously evolved by Queen Mary’s engineering team. The aim of QMOne is to improve efficiency and effectiveness of existing processes, improve staff and students experiences. Currently delivering the first two use cases ( Early Wins): the AskQM Bot, as part of QMOne for Students, and the HR Bot, as part of QMone for Staff. PS, Research and Education are currently developing proposals for further QMOne use cases to follow the Early Wins. 

• Personal Productivity AI tools: These are tools such as Claude, Copilot, Perplexity, Gemini that help students and staff be more productive and more efficient in their day to day work and study.  Currently reviewing tools to identify ones which will be useful for different use cases and the governance around this. These will start to be released in the new year. 

• Low-Code AI: A set of tools that enable staff and students to build their own applications, automations, agents and other solutions with limited or no traditional coding. Currently setting up Centre of Excellence for Copilot Studio and Power Platform so teams across the University can safely build their own solutions.  

• Build the Central Data Repository with pilot use cases and updated data governance. Reporting, automation and AI all depend on a shared and trusted data foundation, which the platform will provide. 

• Also supporting a number of pilots in the research and education space such as academic integrity tools and marking tools,

 

Agentic AI: Refers to systems capable of making decisions and acting autonomously based on user prompts, often employing sophisticated reasoning, goal-oriented behaviour, and iterative planning to address complex problems. These tools operate beyond basic question-and-answer functions and can generate outputs or take actions with other systems with minimal human oversight.

Artificial Intelligence or AI:  refers to a branch of computer science dedicated to developing data processing systems that perform functions normally associated with human intelligence, such as reasoning, learning, adaptation, understanding of abstract concepts and knowledge. Artificial Intelligence (AI) technologies may be deployed for a variety of purposes, including but not limited to: task automation (through macros, chatbots, or robotic process automation); content generation (including the creation of images, videos, text, or music); human representation (for example, through deepfakes, synthetic voices, or digital personas); insight extraction (via machine learning models and data analytics); decision-making support (such as optimisation algorithms and decision trees); and human augmentation (including applications involving exoskeletons, avatars, or other assistive systems.

AI Hallucinations: Outputs generated by an AI system that are incorrect, misleading, or not supported by the input data. In text-based AI (such as chatbots), hallucinations occur when the model presents false information as fact. In computer vision systems, hallucinations occur when the AI incorrectly identifies or interprets objects, people, or features in images or videos.

AI Tool(s): refer to all software applications, technologies, solutions, systems, modules, products and services that use (wholly or partially) AI algorithms to perform tasks associated with human intelligence. These include tools, models, data-processing, codes and other intelligent automation computational systems. The list is non-exhaustive. Note that most of our digital tools now incorporate or leverage AI in some form.

Data owner: A data owner is either a person, or a department that has authority over a dataset, namely on how this data is collected, stored, shared and used. The Data owner has the power to decide who can access or use the dataset and under what conditions. They must ensure that any data used complies with Queen Mary’s policies, data protection legislation (GDPR), and ethical standards. This includes verifying that the data is accurate, anonymised, when necessary, free from bias and legally obtained. Data owners are also accountable for curating datasets in a way that they do not perpetuate harm, reinforce stereotypes or compromise the rights of people from whom the data was derived. Proper documentation, risk assessment, and impact evaluations should accompany any dataset intended to use in AI development.

Generative AI (also known as GenAI): AI that creates content, such as text, images, audio, video or code, in response to instructions.

IDEAS forum/process: The IDEAS Forum is composed of experts from across all areas of IT Services and serves as a consultative body to evaluate proposed ideas and solutions. Its primary role is to review new AI-related initiatives, approve purchase of AI tools provide guidance, and recommend appropriate next steps. The IDEAS Forum does not hold responsibility for the implementation or operational delivery of these services.

Institutional AI Systems: An AI system developed or operated by Queen Mary as part of a university service or process. 

Unmanaged AI Tools: Individuals using AI to assist their own work or study, whether institutionally licensed or personally accessed. QMUL doesn’t control the design, training, behaviour, or outputs of the AI tool. For example: Generative AI assistants (text, image, video, code), Research, summarisation and analysis tools, Learning and study support tools