Governing AI in independent schools: who is accountable and what does good governance look like?

Artificial intelligence is already embedded in school life. It can support lesson delivery, provide wider access to curriculum resources, and enable more personalised learning. It can also make school administration, communications and presentation preparation quicker and more efficient.

Alongside education-specific technology, pupils, staff, governors and parents have access to widely available generative AI tools. The challenge for independent schools is therefore not simply whether AI should be used, but how to provide an effective framework for its responsible use.

In our recent webinar, partners Gemma Woodhouse and Joanne Vengadesan considered the regulatory framework and practical governance measures schools should have in place. You can view the webinar recording here, and this article summarises the key discussion points below.

A holistic regulatory framework

Unlike the EU, the UK does not currently have specific AI legislation or an AI act. Instead, schools need to consider an overarching principles-based framework when deploying and using AI. The key principles are:

  • safety, security and robustness;
  • transparency and explainability;
  • fairness;
  • accountability and governance; and
  • contestability and redress.

These principles sit alongside the data protection laws, equality legislation, and safeguarding and governance obligations, as well as Department for Education guidance. They should not be treated in isolation and schools need to consider their combined effect whenever AI is introduced or used.

Understanding where AI is already being used

Schools should carry out an audit to identify the tools being used by pupils, staff and governors, as well as AI functionality embedded within existing technology. A long-standing supplier may introduce AI into its product without the school having made a separate decision to procure an AI system. The audit should pay particular attention to higher-risk uses involving pupil or staff data, safeguarding, admissions, assessments or other potentially significant decisions.

This oversight is also essential at governor level. Governors need to understand the extent to which AI is being used, the risks it presents and how the risks are being managed. They must also be clear about which AI tools are appropriate for use in their governance role. Access to a sophisticated AI system through a governor’s commercial work, for example, does not necessarily mean it is appropriate for handling school information.

Putting policies and training into practice

Schools should establish clear parameters around approved and prohibited uses of AI.

For staff, the appropriate tool will depend on the purpose. Using AI to prepare a general teaching presentation will have a different risk profile from entering information about a particular pupil to seek assistance with an individual issue. Staff should understand which tools are approved and what information must not be placed into public or unprotected systems.

Pupils also need age-appropriate guidance on acceptable use, including the consequences of passing off AI-generated work as their own. Schools should retain evidence that this guidance and training have been provided.

Consideration should also be given to parents, with schools making it clear that school-related information, including personal or confidential information about pupils and staff, should not be entered into public or unprotected AI tools. This could be addressed through the parent contract or a supplementary policy.

Reviewing suppliers and contracts

Existing technology procurement and data protection requirements continue to apply to AI products, but AI introduces additional considerations.

Schools should consider:

  • what the product does and the purposes for which it will be used;
  • whether the supplier will use submitted content for AI training;
  • if training functionality can be switched off;
  • whether the supplier describes itself as a controller for any of the information;
  • whether international data transfers are involved; and
  • if transparency and privacy notices need to be updated.

Permission to use content for training may not appear in an obvious AI clause. It may be contained in a broader licence or a provision allowing information to be used to improve the service. Supplier terms and privacy information therefore require careful examination.

Where limited training within a closed system is intended to provide a benefit, such as personalised learning, that use should be expressly considered and documented.

Safeguarding and AI

Schools should assess AI risks holistically. Alongside data protection obligations, online safety, security, equality, transparency, bias, and the potential consequences for pupils and staff, should all be considered.

It is also important that AI policies do not direct escalation solely to the IT team and data protection officer. The designated safeguarding lead should also form part of the escalation process, as they may identify safeguarding risks that might otherwise be overlooked.

Preserving human judgement in decisions and parent complaints

Schools are seeing a growing number of parent complaints being drafted with AI which can often be very long in length. Schools may consider asking complainants to provide a concise summary of their concerns so that the focus remains on understanding and resolving the underlying issues.

AI must not, however, become the decision-maker. Asking a system what decision should be made raises serious questions about transparency, human oversight and whether the responsible individual has genuinely exercised their own judgement.

If a decision is challenged, the school may need to demonstrate that the decision-maker considered the relevant information and reached a reasonable conclusion. ‘AI told me what to do’ is not a substitute for that decision-making process.

Schools should also be aware that prompts and outputs may create records capable of disclosure. If litigation is anticipated, which is common with parent complaints, prompts asking an AI system how the school should respond may potentially be disclosable in subsequent proceedings. Prompt and chat histories held by the school may also be relevant to a subject access request.

Practical steps

  • Audit the AI tools and functionality already in use.
  • Identify higher-risk uses involving personal data, safeguarding or significant decisions.
  • Review supplier terms, privacy information and international data transfers.
  • Conduct detailed, context-specific risk and impact assessments.
  • Set clear policies for pupils, staff, governors and parents.
  • Provide and record appropriate training.
  • Establish reporting and safeguarding escalation routes.
  • Retain meaningful human oversight over every significant decision.

Related expertise