Our Approach
GCSE Python Lab uses AI to reduce teacher workload and give students instant feedback. The guiding principle is simple: AI should support learning, not replace teaching.
This page covers how we use AI, what safeguards are in place, and how we comply with UK education regulations.
How We Use AI
1. Student Hints & Guidance
When students get stuck, they can request hints. The AI will ask questions and suggest ways to think about the problem, but it won't give complete solutions or write full code for them.
2. Automated Marking
AI applies mark schemes to coding exercises and written answers to give instant feedback. Teachers can review and override any marks they disagree with.
3. Question Generation
Teachers can use AI to generate question ideas and mark schemes, which they then review and edit before assigning to students.
AI in Content & Platform Development
Question Bank Creation
I used AI to help create the question bank (900+ GCSE and 500+ A Level questions), which significantly sped up content development. Questions were generated based on my own teaching materials from 10+ years in the classroom, with me reviewing everything.
Important: I didn't train AI on OCR past papers, mark schemes, or specification text. I didn't copy and paste from copyrighted exam board materials. The questions are original content aligned to the J277 specification. It's the same approach as writing my own teaching resources, just faster with AI help.
However: I can't guarantee that the underlying AI models (GPT-4) weren't exposed to exam board content during their training by OpenAI. With 900+ GCSE and 500+ A Level questions, cross-checking every single one against all exam board materials isn't practical. All questions are OCR-style but not actual OCR questions. They're original works created with AI assistance. If you have copyright concerns about a specific question, contact me and I'll review it.
Website Development
I'm a Computer Science teacher with a degree in Mathematics and Computer Science. I built this platform using Node.js/TypeScript, Google Firestore, and OpenAI. The development process used AI-assisted code generation tools like GitHub Copilot and Claude.
AI assistants wrote significant chunks of the code, particularly UI components and boilerplate. But the system architecture, feature design, and pedagogical decisions are all mine. As a full-time teacher with limited spare time, using AI allowed me to build something production-quality that would otherwise have taken months of full-time work.
That said: This has still been a huge time investment. Development has been iterative over many months, with extensive testing, vetting of large question banks, and building custom educational tools like the flowchart interpreter and logic gates simulator. AI sped things up, but this is still a substantial project that requires pedagogical expertise and careful quality control.
Safety Mechanisms
System Prompts
Every AI interaction includes system-level instructions that strongly discourage revealing full solutions or writing complete code. It's not foolproof, but it does significantly reduce the chance of students getting complete answers handed to them.
Rate Limiting
There are daily limits on AI hints per student and per-submission feedback limits. AI is completely disabled during test mode to prevent over-reliance.
Teacher Control
Teachers can review and override all AI marks, add their own feedback, disable AI for specific assignments, and control when marks are released.
School Controls
A school chooses which AI features its pupils can use, feature by feature: marking of written answers, guided help buttons, the open tutor chat, code feedback, pupils generating material, and practice hints. Each can be switched off independently, and a school can switch the lot off.
Where AI marking is off, written answers are queued for teachers to mark by hand, and questions marked by ordinary code (multiple choice, matching, binary, boolean, trace tables and similar) carry on marking instantly. A teacher can narrow the school's choice further for a class or a single assignment, and can make an exception for an individual pupil — but nothing below the school can switch on something the school has switched off.
Data Protection
Only question text and submissions are sent to AI. No personal information is shared. OpenAI doesn't use our data for training. Schools that switch AI off send nothing to any AI service at all. See our Privacy Policy.
Safeguarding
Some of our AI features let a pupil type freely. Any tool that does that has to assume a child will one day type something that has nothing to do with schoolwork, and be built for that day rather than surprised by it. This is what happens when they do.
Everything a pupil types is screened
Before a pupil's message reaches the tutor, it is checked by a moderation service for self-harm, violence, sexual content, harassment and hate. What the AI writes back is checked the same way before the pupil sees it. Screening runs on every free-text surface: the tutor chat, the editor help panels and the diagram generator.
A pupil in distress gets an answer, not a wall
If a pupil writes something that suggests they are struggling, they do not get a blank refusal. They get a short, warm reply that does not try to counsel them, names an adult at their school to talk to, and gives them the Childline number. A concern is recorded at the same time. Only sexual content is refused outright.
Somebody at the school is told
Each school names a safeguarding lead contact. Concerns are emailed to them and appear in a queue on their school dashboard, where they can be acknowledged, assigned, noted and closed. The email contains no pupil name and none of what the pupil wrote — that stays behind the school's login.
Pupils are told their teacher can see it
The chat panel says so, plainly, next to the box. The AI is instructed never to tell a pupil their messages are private and never to promise to keep anything secret. It is also instructed to say it is a study tool, not a person, if a pupil asks.
Schools can see what their pupils are doing
Conversations are kept for a period the school chooses, between three months and three years. A school's safeguarding lead contact and admins can read them; class teachers can read their own classes' unless the school restricts that. Schools can print a record of their settings, how those settings have changed, and how concerns were handled — for governors, or for their own filtering and monitoring records.
None of this replaces a school's own safeguarding arrangements, and it is not a monitoring system in the sense that term is usually used — it watches what pupils type to our AI, and nothing else they do on their device. It is one more place a disclosure might surface, built so that it reaches a person when it does.
UK Compliance
I comply with UK Department for Education guidance on AI in education and UK GDPR requirements:
- Teachers maintain control over all AI-generated content
- AI use is clearly labeled and transparent
- Data minimisation: only necessary data is processed
- No personal information sent to AI systems
- Parental consent required for students under 13
See the Privacy Policy for full details.
Limitations
AI marking can make mistakes. You should review the marks for any formal assessments before releasing them.
Students might try to game the hint system. That's why AI features work best for homework and practice, not high-stakes tests.
AI feedback doesn't have the nuance that you bring as a teacher. Think of it as a starting point rather than a replacement for your own guidance.
Best Practices
Recommended:
- Use AI marking for homework and practice work
- Review AI marks before releasing formal assessments
- Disable AI features during tests and mock exams
Avoid:
- Relying solely on AI marks for official grades
- Using AI-generated content without review
My Commitment
I'm committed to being transparent about how AI is used, making sure teachers stay in control, protecting student welfare, and complying with UK education regulations.
Got questions or concerns? Drop me an email: enquiries@forgeclass.co.uk