On the 19th January 2026, the Department for Education published an update to its Generative AI Product Safety Standards, introducing four additional standards and further clarifying expectations for how AI products used in education should be designed, governed and deployed.
Since the initial publication of the standards, Teachmate’s CEO, Jon Chippindall, and CTO, Ian Cunningham, have worked with the DfE to support the development and refinement of these expectations, drawing on their experience of building a teacher-facing AI platform used by over 380,000 educators and 2300 schools.
As a welcome update, the table below sets out how Teachmate meets each of the DfE Generative AI Product Safety Standards, reflecting our ongoing commitment to the safe, responsible and transparent use of AI in education.
| DfE Standard: | How Teachmate Complies: |
| Filtering | These safety expectations only apply to learner-facing platforms. As Teachmate is a teacher-facing service, they do not apply. Despite this, Teachmate continues to use AI models with built-in filtering to mitigate harmful content categories, alongside additional second-pass filtering to ensure high levels of accuracy and educational relevance. |
| Monitoring and Reporting | Similarly to the above standard, this expectation applies only to learner-facing platforms. However, Teachmate logs outputs (only for 28 days) through the ‘My Content’ history feature, supporting transparency and internal monitoring, including the use of detailed internal performance metrics. |
| Security | As Teachmate is exclusively a teacher-facing platform, deliberate attempts at jailbreaking are unlikely. Nevertheless, Teachmate uses a range of technical measures to protect against the jailbreaking and reprogramming of platform functionality. School administrators are assigned elevated permission levels, enabling them to manage user access by assigning and unassigning licences. Teachmate is Cyber Essentials Certified. |
| Privacy and Data Protection | Teachmate’s privacy notice, which covers all required information, is available here: https://teachmateai.com/privacy-policy. Additional information relating to the completion of DPIAs, schools’ duties as data controllers, data protection legislation and AI model training (which Teachmate does not undertake) can be found here: https://teachmateai.com/data-processing-agreement and https://teachmateai.com/data-protection. Further information can also be provided on request. |
| Intellectual Property | Teachmate does not collect, store or share any user data for commercial purposes, including model training (such as fine-tuning), product improvement or product development. The end user of the AI tool retains ownership of the intellectual property of any AI-generated content, in this case the teacher. |
| Design and Testing | Teachmate has a rigorous testing and user acceptance process for its generative AI tools to ensure high-quality outputs for educators. Testing is carried out by both teachers and specialist user groups, with tools only released following a quality assurance sign-off. As generative AI models evolve over time, output quality is monitored and tools are reviewed regularly to maintain high standards. |
| Governance | Formal mechanisms for raising concerns or lodging complaints are set out in Teachmate’s privacy policy: https://teachmateai.com/privacy-policy. Teachmate conducts periodic risk assessments of AI model changes, new feature launches and tool additions, considering both risks to teachers and students as well as technical considerations. |
| Cognitive Development (New for January 2026) | This standard only applies to learner-facing products and does not apply to Teachmate. However, the platform has been specifically designed to support and work alongside teachers’ professional judgement and pedagogical decision-making. |
| Emotional and Social Development (New for January 2026) | This standard only applies to learner-facing products and does not apply to Teachmate. The platform does not interact directly with learners and avoids anthropomorphic or emotionally persuasive design, ensuring AI is used as a professional support tool rather than a substitute for human interaction. |
| Mental Health (New for January 2026) | This standard only applies to learner-facing products and does not apply to Teachmate. The platform does not interact directly with learners and is not designed to detect or respond to mental health concerns, ensuring appropriate boundaries are maintained between professional AI support tools and student wellbeing support. |
| Manipulation (New for January 2026) | Teachmate does not employ any tactics or techniques designed to encourage users to remain on the platform for longer, create a sense of social pressure or manipulate emotional responses in order to influence user behaviour. The platform does not use dark patterns and does not incorporate advertising or promotional material. |