Safety & AI: Is AI Tutoring Safe for Students?

If you're a parent or school leader reading this, the fact that you're asking whether an AI learning tool is safe is a good thing. It means you care about the young people in your life, and frankly, you should be asking these questions of every technology your children use.

AI in education is still new. The rules are still forming. And the honest truth is that not every company building in this space has earned your trust yet. We think scepticism is healthy, and we'd rather answer hard questions openly.

This article covers the three pillars of how we approach safety: data protection, content accuracy, and responsible AI. We'll be transparent about what we do well, honest about where we're still improving, and clear about how we compare to general-purpose AI tools like ChatGPT.

Our Commitment to Safety

Safety was something that was planned for, it was part of the foundation to how Medly was designed.

When you've spent years in clinical practice, you learn that safety isn't something you can retrofit. You build it into systems from the ground up, test it relentlessly, and never assume it's finished. That's the approach our founders brought to Medly.

Every feature on the platform, from the AI tutor to the marking engine to the way we handle data, was built with safeguarding in mind. We don't collect data we don't need. We don't let the AI operate without guardrails. When something isn't right, we fix it and tell you about it.

This isn't a legal obligation we grudgingly comply with. If it wouldn't be safe enough for our own families, it's not safe enough for yours.

Data Protection and Privacy

Data privacy is one of the first things parents ask about, and rightly so. Here's exactly what we do and, just as importantly, what we don't do.

Medly is fully GDPR compliant and registered with the Information Commissioner's Office (ICO) under registration number ZB724661. You can verify this yourself on the ICO's public register.

What we collect (and what we don't)

We follow a strict data minimisation principle. For regular users, we don't collect your child's name, home address, or school. We collect what's needed to make the platform work and nothing more. This approach aligns with the ICO's Age Appropriate Design Code, which sets out standards specifically designed to protect children online.

Your child's data is not used to train third-party AI models

This is a critical point. When your child uses Medly, their conversations, answers, and learning data are not sent to OpenAI, Google, or Anthropic to train their models. We have explicit agreements in place to prevent this. We may in the future use anonymised data to improve Medly's own inherent model capability, but we will notify you in advance and can opt out of that too.

Payment security

If you subscribe through the Apple App Store, payment is handled entirely by Apple. Medly never sees, stores, or processes your card details. For web-based subscriptions, we use an industry-standard third party secure payment processor called Stripe.

Where your data lives

Our infrastructure runs on data centres in the UK and US, with data stored closest to where you set up your account. This means your data is stored securely and locally.

Your rights

You have the right to request deletion of your data at any time. If you email us, we'll action it. Our privacy policy sets out everything in full detail.

Content Accuracy

An AI tutor is only useful if you can trust what it teaches. We take content accuracy seriously, and we're honest about where we are in that journey.

Teacher-reviewed content

Every piece of content on Medly is reviewed by experienced, qualified teachers. Our reviewers have upto and in some cases over 30 years of teaching experience. They hold Qualified Teacher Status (QTS). Many are senior examiners for major exam boards. Several are former Heads of Department with postgraduate qualifications. They bring real-world curriculum specific expertise.

Aligned to your exam board

Medly content isn't generic "GCSE revision." It's built to match the specific specification of your exam board. If your child is studying AQA Biology, they get AQA Biology content, not a vague summary that might cover the right topics.

Being honest about AI limitations

Here's where we're going to be straight with you. AI is not perfect. No AI system, including ours, gets everything right 100% of the time.

What we can tell you is that Medly has a significantly lower error rate than general-purpose AI models like ChatGPT for curriculum-specific tasks. That's because our content goes through teacher review, our AI is constrained to specific exam board specifications, and we've built subject-specific models rather than relying on one general-purpose system.

We're also actively rebuilding our marking engine to further reduce error rates, these are genuinely hard problems, and we'd rather be honest about the challenge than pretend we've completely solved it.

Medly is not a replacement for a professional teacher. It's a powerful study tool that complements classroom learning, provides unlimited practice, and gives instant feedback. But the best outcomes happen when students use Medly alongside good teachers, not instead of them.

Medly vs ChatGPT: A Safety Comparison

A lot of students tell us they've tried using ChatGPT for revision. We understand why. It's free, it's powerful, and it can answer almost any question. But there's a significant difference between a general-purpose AI tool and one specifically built for learning.

The key difference comes down to purpose. ChatGPT is a general tool built for everyone. Medly is specifically designed for learning, built for students, and reviewed by teachers. When it comes to exam preparation, that distinction matters.

Safeguarding

Young people sometimes share things with technology that they wouldn't say to a person. We take that responsibility seriously.

Our safeguarding team

Medly has designated safeguarding leads who are enhanced DBS checked. We audit and regularly test our safeguarding protocols and lockouts, with humans in the loop on escalated disclosures.

AI-powered safety monitoring

Every conversation on Medly passes through an AI-powered safety layer. The system uses contextual understanding to detect both direct and indirect expressions of distress. A student doesn't need to use a specific word for the system to recognise that something is wrong.

What the safety system detects

Our safeguarding AI is trained to identify:

• Suicidal ideation or self-harm
• Abuse, neglect, or exploitation disclosures
• Domestic violence indicators
• Significant distress or hopelessness
• Severe bullying
• Substance misuse
• Exam anxiety, peer pressure, or family-related stress

How Medly responds

When the system detects a concern, the response follows a clear protocol. Medly listens with empathy and validates what the student is feeling. It is honest about its limitations, clearly stating that it is not a counsellor and cannot provide professional support. It provides crisis helpline information immediately. It encourages the student to speak to a trusted adult. Crucially, it pauses the tutoring session. Medly will not carry on with revision as if nothing has happened.

The AI never attempts to counsel the student, never probes for details, and never minimises what they've shared. In certain situations, parents may be notified by email that a concern has been detected, without sharing the specific details of the conversation, and never if doing so could put the child at risk.

Crisis resources

If you or a young person you know needs support right now:

• Childline: 0800 1111 (free, 24/7)
• Shout Crisis Text Line: Text SHOUT to 85258
• Samaritans: 116 123 (free, 24/7)
• YoungMinds: Text YM to 85258

You can also contact our safeguarding team directly with any questions at safeguarding@medlyai.com.

Transparency

We believe transparency builds trust, and we'd rather show you how things work than ask you to take our word for it.

Open about our AI architecture

Medly doesn't rely on a single AI provider. We use a multi-model architecture combining models from all major providers with a proprietary model router that selects the best model for each task. This means we're not locked into one provider's limitations, and we can adapt as the technology improves.

Continuous testing

We regularly test our safeguarding responses to make sure they work as intended. This includes simulated scenarios reviewed by our safeguarding team to identify gaps and improve detection accuracy.

Frequently Asked Questions

Does Medly sell student data?

No. Medly does not sell, share, or trade student data with any third party. Your child's data is used solely to provide and improve the Medly learning experience. We don't run advertising, so there is no commercial incentive to monetise user data.

Is Medly GDPR compliant?

Yes. Medly AI Ltd is fully GDPR compliant and registered with the Information Commissioner's Office (ICO) under registration number ZB724661. We comply with the Age Appropriate Design Code, the Online Safety Act 2023, and the UK Department for Education's guidance on generative AI product safety.

Can Medly detect if a student is in distress?

Yes. Medly has an AI-powered safety layer that monitors conversations for signs of distress, including suicidal ideation, self-harm, abuse, bullying, and significant anxiety. The system uses contextual understanding rather than simple keyword matching, so it can recognise indirect expressions of distress. When a concern is detected, Medly responds with empathy, provides crisis resources, and pauses the tutoring session.

Is Medly's content accurate?

Medly's content is reviewed by experienced teachers, many of whom are senior examiners for major UK exam boards. All content is aligned to specific exam board specifications. However, we're honest that no AI system is perfect. Medly has a lower error rate than general-purpose AI models for curriculum-specific tasks, and we are continuously improving, including actively rebuilding our marking engine for complex tasks like English essays and Maths proofs.

What happens if Medly gets something wrong?

If you spot an error, we want to know about it. You can report issues through the platform or email us directly at contact@medlyai.com. We investigate every report and correct errors promptly.

Is Medly appropriate for under-16s?

Yes. Medly is designed for secondary school students, including those under 16. We comply with the ICO's Age Appropriate Design Code, which sets specific standards for services likely to be accessed by children. We apply data minimisation principles, don't collect unnecessary personal information, and have built safeguarding systems specifically designed to protect younger users.

---

Safety in education technology isn't something that gets finished. It's something you commit to, invest in, and improve every single day. We're proud of the systems we've built, and we're honest about the work still ahead.

If you'd like to learn more, you can read our full privacy policy, visit our safety page, or read more about what Medly is and how it works. If you have a specific safeguarding concern, you can reach our team at safeguarding@medlyai.com.

We built Medly for students who deserve better access to education and keeping them safe while they learn is imperative.