Navigating the pitfalls of AI in Educational Psychology

Artificial Intelligence (AI) has already become a fixture within many educational settings. The ease with which it can be used to write lesson plans, coupled with the promise of enhanced analysis and streamlined reporting, makes AI a very attractive proposition for educators.

However, for Educational Psychologists (EPs) the wider adoption of AI tools comes with considerable risks. Through the use of AI, the law around data privacy and security can easily be overlooked, particularly by those keen to take a learn-by-doing approach rather than seeking a deeper understanding before diving in.

Child data is not safe with AI

One of the most critical pitfalls when using AI platforms is the exposure of sensitive data. Many free or basic tools retain user-submitted content indefinitely to improve their algorithms. This means that any child data processed — even if anonymised — could be absorbed into AI platforms’ datasets. Transparency about AI tools’ data handling is often limited, and users have little control over how long their information is stored, what country it’s stored in, or who can access it. Once absorbed into supporting datasets information shared cannot be removed and will be used to create, and may even be cited, in AI responses to other users’ enquiries.

Only the far more expensive, higher-tier, enterprise-level subscriptions offer robust privacy controls, data deletion, and contractual assurances that user data will not be harvested or used to further train AI models. Without such protections, education professionals dipping their toes into AI through lower-cost alternatives will be likely submitting confidential information into systems that repurpose it. This can only compromise the privacy and safety of vulnerable children.

This is not me being over-anxious, the risk is especially pronounced for EPs. Educational Psychology reports are by design data and information rich; frequently containing a large amount of sensitive personal information, including developmental histories, psychological assessments, behavioural observations, and recommendations. The inadvertent exposure of such data to AI tools could lead to breaches of GDPR.

Anonymisation is a placebo for data security

A frequent misconception is that simply anonymising reports and using AI piecemeal is sufficient to protect privacy. While removing names of children and settings may seem like an obvious safeguard, we must remember that AI systems are built to uncover patterns and make connections within data. AI algorithms can infer identities, reconstruct sensitive information from seemingly innocuous details, or even create complete reports by cross-referencing previously submitted information. A report that omits a child's name but includes specific behavioural incidents, school year, and intervention history could still be linked to an individual, especially within specialised educational settings or a relatively small client base. Anonymisation alone provides only a superficial layer of protection against the sophistication of AI-driven algorithms.

AI can write a report but can’t do your thinking

Assuming the use of enterprise-level subscriptions, using AI to draft reports and analyse assessment data is reasonable enough. However, using AI to generate recommendations and interventions without proper scrutiny on the part of the EP is deeply problematic. Sadly, as professionals we cannot trust the apparent objectivity and efficiency of the technology. AI models are only trained on existing data, which is unlikely to represent the diversity or complexity of the children with who EPs’ work. This can only result in the propagation of biases and misinterpretations and will directly impact the quality of psychological intervention and support.

“Copy & Paste” does not make you legally compliant

The UK regulatory landscape around AI is clunky compared to that in the EU but it is evolving. EPs must stay abreast of changes in data protection law, and ethical best practice. The Information Commissioner's Office (ICO) has issued guidance on the use of AI, emphasising the need for explicit consent, transparency, and accountability. Some pay-for legal support for EPs — even around new documentation requirements — is being provided by a selection of specialist education sector solicitors. However, if the EPs themselves have no deeper understanding of AI, copying and pasting pre-written passages into key documentation does not mean that those EPs’ understand the risks of AI, that their practice is ethically sound, or even compliant with the law.

Use AI, but with caution

AI has the potential to enhance the work of Educational Psychologists, but its use demands critical thinking, and a deep awareness of data security risks. EPs must take advice from legal advisors, and professional bodies before integrating AI into their workflows. Training in digital literacy and AI ethics are essential, and robust policies must be developed for privacy and consent, with separate data-sharing agreements put in place to reflect the latest regulatory guidance. Educational Psychologists, like all professionals, need to engage with AI responsibly, ensuring that technological innovation does not come at the expense of professional integrity.