How to Choose the Right AI Tools for Teaching: A Practical Framework for Universities

Why Universities Can’t Afford to Wing It

AI adoption in higher education isn’t optional anymore. Students are already using these tools. Employers expect graduates to be AI-literate. The question isn’t whether to integrate AI — it’s how to do it without creating problems you’ll spend years cleaning up.

The core risks fall into three buckets: data privacy, ethical misuse, and pedagogical harm. A tool that leaks personally identifiable information, trains its model on your students’ work, or produces outputs educators accept uncritically can damage trust, violate regulations, and produce graduates who can’t think critically about AI.

Getting this right starts with governance, not tools.

Step 1: Build a Governance Committee Before You Pick a Single Tool

Before you evaluate any platform, establish who makes the decisions. This isn’t bureaucracy for its own sake — it’s how you avoid one enthusiastic lecturer accidentally committing your institution to a tool with a GDPR nightmare buried in its privacy policy.

A strong AI tool governance committee should include representatives from:

• Information security — to assess technical vulnerabilities
• Data protection — to evaluate compliance with GDPR and institutional policy
• Pedagogy — to determine whether the tool actually supports learning outcomes
• Legal or compliance — to review end-user licence agreements
• Subject specialists — to assess disciplinary fit

The University of Chester built exactly this kind of cross-functional committee early in their AI journey. Their guiding principle: cautiously, transparently, and responsibly embrace AI — balancing genuine benefits against security and data protection risks. That principle-first approach is what keeps governance from becoming a rubber stamp.

Step 2: Create an Approved Tools List — and Stick to It

Your institution needs a curated list of tools that have passed your governance review. This isn’t about limiting innovation. It’s about ensuring that when educators use AI in their teaching, they’re working within a framework that protects everyone involved.

The University of Chester’s approved list includes Microsoft 365 Copilot, TeacherMatic, Adobe Firefly, CareerSet, and Shortlist.me. Each of these passed scrutiny on data handling, security, and pedagogical value.

When building your own list, prioritise tools that offer at least one of the following:

• EU data storage — ensuring GDPR compliance by design (e.g., CareerSet)
• No model training on user data — your inputs stay yours (e.g., Microsoft 365 Copilot for Enterprise)
• Time-limited data retention — inputs and outputs aren’t stored indefinitely (e.g., TeacherMatic)
• Multifactor authentication — basic but non-negotiable security hygiene

These aren’t nice-to-haves. They’re the baseline for responsible institutional use.

Step 3: Use a Framework to Evaluate Tools Not Yet on the List

Here’s the reality: educators will always find tools that aren’t on the approved list but seem perfect for their course. You need a process for handling that — one that doesn’t just say “no” by default.

The University of Chester adapted Charles Sturt University’s S.E.C.U.R.E. framework into their own SafeAI framework. It uses a structured questionnaire to guide educators through a self-assessment before using an unapproved tool in a specific context.

The SafeAI Questionnaire: Key Questions to Ask

1. What data are you entering into this tool?

Would you be inputting any of the following?

• Personally identifiable information (student names, IDs, contact details)
• Confidential institutional data
• Security credentials (usernames, passwords)
• Copyrighted material (paywalled journal articles, licensed content)

If yes to any of these, stop. The tool is not appropriate for that use case without further review.

2. Are you using the tool for anything ethically questionable?

This includes generating content designed to mislead, automating assessments in ways that bypass academic integrity policies, or using outputs to make high-stakes decisions without human review.

3. Will you evaluate the outputs critically?

Using AI outputs without scrutiny isn’t just bad practice — it’s a pedagogical failure. If the answer is “I’ll just use whatever it produces,” the tool isn’t the problem. The process is.

If an educator answers “no” to all three concern areas, they can proceed with using the tool in that specific context. This keeps the framework practical without abandoning oversight.

Step 4: Read the Fine Print — Seriously

End-user licence agreements and privacy policies are not optional reading. They are the legal contract between your institution and the tool provider. Most educators skip them. That’s a mistake.

Two things to look for immediately:

Data training clauses — Does the tool use your inputs to improve its model? Many free tools do. If your students’ work is being used to train a commercial AI, that raises serious ethical and potentially legal questions. Most major chat-based tools offer an opt-out — find it and use it.

Free tools aren’t free — When a tool costs nothing, your data is usually the price. That’s not inherently disqualifying, but it needs to be understood and factored into your decision. A free tool that trains on student data is a different proposition than a paid enterprise tool with contractual data protections.

A useful rule of thumb: if you can’t find a clear privacy policy and data retention statement on the tool’s website within five minutes, treat that as a red flag.

Step 5: Embed AI Literacy Into Your Teaching — Not Just Your Tools

Choosing the right tools is only half the job. The other half is teaching students how to use AI responsibly, not just that they should use it.

This means modelling ethical use in your own practice. If you use AI to draft course materials, say so. If you use it to generate feedback templates, explain how you reviewed and adapted the output. Transparency isn’t weakness — it’s exactly the behaviour you want students to replicate in their careers.

Step 6: Keep Reviewing — The Ecosystem Doesn’t Stand Still

A tool that passed your governance review in 2024 may have changed its data practices by 2026. Privacy policies get updated. Ownership changes. New features introduce new risks.

Build a regular review cycle into your governance process — at minimum annually, ideally every six months for tools in active use. Assign someone to monitor significant changes to the tools on your approved list.

This isn’t paranoia. It’s basic institutional hygiene in a fast-moving ecosystem.

A Quick Decision Checklist for Educators

Before using any AI tool in your teaching, run through this:

• Is the tool on your institution’s approved list?
• If not, have you completed your institution’s self-assessment framework (e.g., SafeAI)?
• Have you checked the privacy policy and data retention terms?
• Are you entering any personally identifiable, confidential, or copyrighted data?
• Will you critically evaluate all outputs before using them?
• Are you prepared to be transparent with students about your AI use?

If you can answer these confidently, you’re in a strong position to proceed responsibly.

The Bottom Line

Choosing AI tools for teaching isn’t about finding the most impressive platform. It’s about finding tools that are secure, compliant, pedagogically sound, and aligned with your institution’s values — then using them in ways that model exactly the critical, ethical engagement you want your students to develop.

The universities that get this right won’t just protect themselves from risk. They’ll produce graduates who understand AI deeply enough to use it well, question it when necessary, and lead in a world where that distinction matters enormously.

Start with governance. Build your framework. Then choose your tools. In that order.