How Third Space Learning Safeguard Your Students With AI tutoring

AI tutoring brings new opportunities to support more students than ever before. But it also raises important safeguarding questions. As school leaders, you need to know that any AI tutoring solution meets the same rigorous safeguarding standards you’d expect from human tutors.

At Third Space Learning, we’ve applied our decade of safeguarding experience to ensure Skye, our spoken AI maths tutor, keeps students safe while they learn. This means comprehensive safeguards built into the system from the ground up, not just added as an afterthought.

Built-in safeguarding from day one

Unlike general AI tools that schools might adapt for education, Skye has been designed specifically for school environments with our safeguarding policy as a core priority.

The first level of protection we have is ensuring that we do not recruit anyone, at any level in the company, who may pose a danger to children, following the guidance listed in Section 3 of Keeping Children Safe in Education. 

Here’s how we’ve ensured student safety with AI tutor Skye:

Controlled content and interactions

Lessons created by qualified teachers: Skye can only access learning content created by our team of current and former teachers. The AI cannot generate its own lessons or display anything beyond these specially-designed materials, which are fully aligned with the UK National Curriculum.

Strict interaction boundaries: Skye operates within carefully defined guardrails that control the scope of all interactions. If a student attempts to steer the conversation away from maths, Skye politely redirects them back to the lesson content.

Shared screen safety: The visual component of each session shows only the lesson slides created by our academic team. Students never see generated content, inappropriate images, or material from external sources.

Comprehensive monitoring and recording

Every session is recorded: All tutoring sessions with Skye are recorded and transcribed, creating a complete audit trail. Schools can request access to any session recording at any time.

Automated safeguarding alerts: Our monitoring system automatically identifies and flags potential safeguarding concerns, categorising them from 1 (most severe) to 5 (least severe):

• Category 1 concerns (safeguarding disclosures) trigger immediate alerts for near real-time review
• Categories 2-5 (including disruptive behaviour, student distress, or frustration) are compiled into daily reports

Human oversight: Every flagged session is reviewed by our UK-based safeguarding team, with Category 1 concerns escalated to your school’s Designated Safeguarding Lead immediately.

When students make disclosures

While AI tutoring removes risks associated with human tutors, students may still make safeguarding disclosures during sessions. Skye has been trained to recognise and respond appropriately to potential disclosures about:

• Physical abuse, neglect, or emotional abuse
• Child-on-child abuse or bullying
• Serious violence or criminal exploitation
• Mental health concerns

Our response process

• Recognise: Skye’s training enables it to identify language patterns that may indicate safeguarding concerns, even when not explicitly stated.
• Respond: The AI responds with empathy while making clear that concerns need to be shared with teachers, following established safeguarding protocols.
• Report: Category 1 concerns are flagged instantly to our safeguarding team, who contact your school’s DSL.

School-first approach to safeguarding

We maintain the same principle that guides traditional tutoring: schools are best placed to investigate and act on safeguarding concerns. Your teaching staff have the relationships, context, and local knowledge needed to effectively support students and provide the child with an encouraging environment in which to talk.

Furthermore, if the school decides the local authority or police need to be involved, then they are likely to already know the best people to contact. 

Our role is to:

• Detect and flag potential concerns quickly and accurately
• Provide you with complete session recordings and transcripts
• Support any follow-up investigations you need to undertake
• Maintain detailed records for child protection purposes

Your school’s DSL remains responsible for deciding what action to take, whether that involves contacting parents, social services, or other agencies.

Data protection built for education

Minimal data collection: We only collect the data necessary for effective tutoring. Typically, just the student’s first name, year group, and learning objectives.

No training on student data: Skye is not trained on any student interactions, conversations, or personal information. The AI’s training is exclusively from our academic team’s input and prompts.

GDPR compliance: All data handling meets UK GDPR requirements, with clear lawful bases for processing and appropriate security measures.

School control: You maintain ownership of all session data and can request deletion at any time.

Addressing common concerns about AI tutoring safeguarding

“Can AI really spot safeguarding concerns as well as humans?”

While AI cannot replace human judgment, Skye’s monitoring system captures every word of every session, something impossible to achieve with traditional tutoring at scale. Our automated flagging system identifies potential concerns that might be missed in busy school environments.

“What if the AI gives inappropriate responses to disclosures?”

Skye has been specifically trained on safeguarding scenarios using established protocols. The responses are consistent, appropriate, and immediately escalated to qualified humans for follow-up.

“How do we know the AI won’t malfunction or behave unpredictably?”

Skye operates within strict technical guardrails that prevent it from accessing external content or deviating from its training. All sessions are monitored, and any unusual behaviour would be immediately flagged.

The safeguarding advantage of AI tutoring

AI tutoring with Skye offers some safeguarding benefits over traditional models:

• Complete transparency: Every interaction is recorded and available for review, providing full visibility into what happens during tutoring.
• Consistent standards: Unlike human tutors who may vary in their safeguarding awareness, Skye applies the same high standards in every session.
• Immediate escalation: Concerns are flagged and escalated within hours, not after weekly supervision meetings.
• No recruitment risks: There are no background checks, references, or ongoing vetting requirements. The safeguarding risks associated with human tutors simply don’t exist.

Proven safeguarding experience

Our safeguarding policies and procedures have been reviewed and approved by the Department for Education as part of the National Tutoring Programme. We’ve applied this decade of experience to ensure AI tutoring maintains the same rigorous standards schools expect.

Since launching Skye, we’ve handled safeguarding concerns with the same professionalism and urgency that schools rely on. Our UK-based safeguarding team includes qualified teachers who understand both the technology and the educational context.

Next steps for your school

If you’re considering AI tutoring, ask potential providers:

• Are all sessions recorded and available for review?
• What automated safeguarding monitoring is in place?
• How quickly are serious concerns escalated to schools?
• What training has the AI received on safeguarding scenarios?
• Who reviews flagged sessions, and what are their qualifications?

At Third Space Learning, we believe AI tutoring should make safeguarding easier, not harder. By building comprehensive protections into Skye from the ground up, we’ve created a system that supports more students safely – giving you confidence in both the educational outcomes and child protection standards.

We always ensure that our safeguarding responsibilities are adhered to and in line with guidance, so that we prioritise the welfare and well-being of children.

Read more: The 12 Most Important Ofsted Safeguarding Questions and Answers [2024]