Adaptibles AI and Automated Decision-Making Policy
Last Updated: Saturday, 14th March 2026
1. Purpose and Scope
This policy explains how Adaptibles CIC uses artificial intelligence (AI) and automated systems across the Adaptibles platform. It sets out what our AI systems do, what they do not do, what safeguards are in place, what rights you have, and how we are accountable for the decisions our systems make or influence.
This policy applies to all users of the Adaptibles platform, including learners, instructors, and visitors. It covers every context in which AI or automation plays a role in shaping your experience or in making decisions that affect you.
We have written this policy in plain language. We believe that people affected by AI systems deserve to understand them without needing a legal or technical background to do so.
2. Our Principles
Adaptibles was built for people who have often been misunderstood, misassessed, or excluded by systems that were not designed with them in mind. We carry that history into how we build and govern our AI. Our AI principles are:
Human-Centred
AI on Adaptibles exists to support human learning and human judgment, not to replace it. Every AI system we deploy is designed to assist, not to determine. The person always matters more than the output.
Transparent
We tell users when AI is involved in their experience. We do not obscure AI-generated content as human-authored, and we do not apply automated decisions without telling the person affected.
Explainable
Where AI makes or contributes to a decision that affects a user, we commit to being able to explain the basis for that decision in plain language. We do not deploy AI systems whose outputs we cannot account for.
Fair and Non-Discriminatory
We actively work to identify and mitigate bias in our AI systems, particularly bias that could disadvantage neurodivergent users, users from minoritised backgrounds, or users whose communication styles differ from dominant norms. We do not treat difference as a signal of risk.
Always Reviewable
No automated decision that significantly affects a user is final without the opportunity for human review and appeal. This is both our policy and, in relevant cases, a legal requirement under UK GDPR.
Privacy-Preserving
Our AI systems are designed with data minimisation in mind. We use the least amount of personal data necessary to achieve the intended function. Special category data, including neurodivergent and disability-related data, is subject to the highest level of protection.
3. AI Systems on the Adaptibles Platform
Adaptibles currently uses AI in two primary areas: Ada, our learning assistant, and our content moderation system. Each is described in detail below.
4. Ada: The Adaptibles Learning Assistant
4.1 What Ada Is
Ada is Adaptibles' AI-powered learning assistant. Ada is a supportive learning companion built into Adaptibles. From guidance and encouragement to personalised reflections and reminders, Ada helps learners move through their experience in a way that feels more accessible, human and less overwhelming. She is also designed to embody warmth, patience, and non-judgement. Ada is available throughout the learning experience to support, guide, and encourage.
Ada operates in two states: a pre-login state, in which it is available to visitors to the platform with limited functionality, and an authenticated state, in which it has access to your learning profile and course history to provide contextually relevant support.
4.2 What Ada Does
In its authenticated state, Ada can:
- Answer questions about course content and concepts
- Offer explanations in different formats to suit your learning style
- Provide encouragement and help you navigate moments of difficulty
- Remind you of your progress and learning goals
- Suggest resources or next steps based on where you are in your learning journey
- Help you reflect on what you have learned
In its pre-login state, Ada can:
- Answer general questions about the Adaptibles platform
- Provide information about neurodivergent learning and the platform's approach
- Guide visitors through the registration process
4.3 What Ada Does Not Do
Ada has defined ethical limits. These are not adjustable by users or operators. Ada will not:
- Provide medical, psychiatric, therapeutic, or clinical advice of any kind
- Offer diagnoses or diagnostic suggestions
- Engage in crisis intervention or attempt to replace professional mental health support
- Make binding decisions about your access to the platform or your course outcomes
- Collect or store information beyond what is necessary for the learning interaction
- Pretend to be a human when sincerely asked whether it is an AI
- Engage with requests that could cause harm to the user or others
- Share your conversation history with other users or with instructors without your explicit consent
4.4 Ada's Limitations
Ada is an AI system. It can make mistakes. It may occasionally misunderstand a question, provide an inaccurate answer, or respond in a way that does not account for your specific context. You should apply your own judgment to Ada's responses and seek professional advice where a matter is serious or complex.
Ada's responses are generated in real time and are not manually reviewed before delivery. We conduct regular quality assessments of Ada's performance and update its parameters where we identify patterns of error or harm.
Ada does not have persistent memory between separate sessions unless your profile data explicitly provides context. It cannot recall details from previous conversations unless those details are stored in your learning profile.
4.5 Ada and Emotional Interactions
Ada is designed to be warm and supportive. However, if a user appears to be in distress or expresses thoughts of self-harm, Ada will not attempt to provide crisis support. Instead, Ada will acknowledge the person, encourage them to seek support from a qualified professional or trusted person, and signpost relevant resources. Ada will also flag the interaction for review by the Adaptibles safeguarding team in accordance with our Safeguarding Policy.
We are transparent about this boundary because we believe it is more respectful and safer than allowing users to form a dependency on an AI system for emotional support in crisis.
4.6 Transparency in Ada's Responses
All responses generated by Ada are clearly attributed to Ada as an AI assistant. Ada will not claim to be human. Where Ada's response draws on specific course material or platform documentation, it will indicate this where possible.
5. Content Moderation System
5.1 Overview
Adaptibles uses an AI-assisted content moderation system to review posts, messages, and other user-generated content published in community spaces on the platform. The system operates across four tiers, each representing a different level of severity and a different response.
5.2 The Moderation Tiers
Tier 1: Automated Review and Temporary Hold
Content that is flagged by the AI system at Tier 1 is temporarily hidden from public view while it is queued for human review. The user who posted the content is notified that their post is under review. No permanent action is taken at Tier 1. The content is held, not removed.
Tier 2: Escalated AI Flag
Content flagged at Tier 2 involves a higher confidence signal from the AI system and is escalated immediately to a human moderator for priority review. The content remains hidden pending that review.
Tier 3: Human Moderator Review
All Tier 1 and Tier 2 flagged content is reviewed by a human moderator. The moderator considers the content in full context, including the user's history, the nature of the community space, and the intent behind the communication. The moderator may reinstate the content, leave it hidden, issue a warning, or escalate to Tier 4.
Tier 4: Senior Review and Formal Action
Cases involving potential legal risk, safeguarding concerns, or serious or repeated violations are escalated to Tier 4 for senior review. Tier 4 may result in account suspension, permanent removal, or referral to relevant authorities.
Auto-Hidden Content
Certain categories of content, including content matching known patterns of hate speech, content that appears to sexualise or endanger minors, or content that appears to constitute an imminent threat, may be auto-hidden immediately without waiting for human review. These categories are defined conservatively and reviewed regularly to prevent over-censorship.
5.3 What the System Does Not Do
Our automated moderation system does not:
- Take permanent action against a user's account without human review
- Flag content solely on the basis of neurodivergent communication styles, including unconventional grammar, direct or blunt phrasing, or atypical sentence structure
- Discriminate against users on the basis of their declared learning profile or accessibility needs
- Apply different moderation standards to different demographic groups
- Share flagged content with other users
5.4 Bias and Neurodivergent Communication
We are specifically aware that automated content moderation systems trained on majority-norm data can incorrectly flag neurodivergent communication styles as problematic. Bluntness, directness, unconventional punctuation, repetition, and non-standard syntax are not violations of our Community Guidelines and should never be treated as such.
We conduct ongoing audits of our moderation system to identify and correct bias. Human moderators are trained in neurodivergent communication norms as part of their onboarding. If you believe your content was flagged because of your communication style rather than its substance, you are encouraged to appeal.
5.5 Notifications
When content you have posted is flagged by our moderation system, you will receive:
- A platform notification informing you that your post is under review
- An email notification with a link to your moderation cases log and a link to our Community Guidelines
- A follow-up notification, by both platform and email, once a human moderator has reviewed your case and a decision has been reached
Notifications will explain what action has been taken and will include information on how to appeal.
6. Course Recommendations and Personalisation
Adaptibles uses your declared learning profile and your course activity to personalise recommendations and adapt your experience on the platform. This includes:
- Recommending courses based on your declared interests and learning goals
- Ordering content in ways that reflect your stated preferences
- Adapting Ada's communication style to your profile
Personalisation on Adaptibles is based on what you tell us about yourself. We do not infer sensitive characteristics from behavioural data or apply profiling that you have not consented to.
You can update your learning profile and preferences at any time from your account settings. Changes take effect immediately.
7. What Counts as an Automated Decision
Under UK GDPR, you have specific rights in relation to decisions made solely by automated means that produce legal or similarly significant effects on you. In the context of Adaptibles, this means:
Decisions That Are Automated
The initial flagging of content by our moderation system is automated. The temporary hiding of content at Tier 1 is automated. These actions are immediate and do not involve a human in the first instance.
Decisions That Are Not Solely Automated
Any action that affects your account status, including warnings, restrictions, or suspensions, requires human review and authorisation. A human moderator must approve every action beyond the initial temporary hold. This means that no significant decision affecting your account is made solely by automated means.
Course recommendations and personalisation features do not constitute significant automated decisions in the legal sense, as they are entirely reversible, do not restrict your access to content, and are based on preferences you have provided.
8. Your Rights in Relation to AI
The Right to Know
You have the right to know when AI is being used in a way that affects you. This policy fulfils that right at a general level. If you want to know whether a specific decision affecting your account involved automated systems, you may ask us using the email below.
The Right to Human Review
If you believe a moderation decision affecting your account was made incorrectly, you have the right to request human review of that decision. All moderation outcomes are subject to human review as a matter of policy, but you may request a fresh review at any time using the email below with the subject line 'Moderation Appeal'.
The Right to Explanation
If a decision has been made about your account, you have the right to a plain-language explanation of the basis for that decision. We commit to providing this within five working days of your request.
The Right to Correct Your Profile Data
If you believe that the data used to personalise your experience is inaccurate, you have the right to correct it. You can update your learning profile directly in your account settings or contact us using the email below.
The Right to Object
You have the right to object to automated processing of your personal data where that processing is based on our legitimate interests. Where you object, we will assess whether there are compelling grounds to continue and will inform you of our decision.
9. Human Oversight and Accountability
Adaptibles maintains the following structures to ensure meaningful human oversight of our AI systems:
- All content moderation decisions beyond a temporary hold require human authorisation
- Ada's response parameters are reviewed and updated on a regular cycle
- Moderation outcomes are audited periodically to identify patterns of error or bias
- Any user may appeal a moderation decision and receive a response from a named human reviewer
- Safeguarding-related interactions flagged by Ada are reviewed by a human team member
- A named internal lead is responsible for AI governance at Adaptibles, accountable to the founder and CEO
We do not treat AI outputs as inherently correct. We treat them as inputs to human judgment.
10. Third-Party AI Providers
Ada and our moderation system are built using third-party AI infrastructure and large language model providers. We select these providers on the basis of their data protection standards, their transparency about model training, and their alignment with our values where possible.
All third-party AI providers processing personal data on behalf of Adaptibles are governed by data processing agreements that comply with UK GDPR requirements. Personal data is not used by third-party providers to train their general models without your explicit consent.
Details of third-party providers involved in AI processing are available on request using the email below.
11. Regulatory Context
Adaptibles operates primarily under UK law, including the UK GDPR and the Data Protection Act 2018. We monitor the development of AI regulation in the UK and internationally, including:
- The UK Government's AI Regulation Policy Paper and the principles-based framework being developed by UK regulators
- The EU AI Act, which, while not directly applicable to UK-based organisations, is shaping international standards and may affect us where we have users in the European Economic Area
- ICO guidance on AI and data protection, including the ICO's Guidance on AI and Data Protection
We commit to updating this policy as the regulatory environment develops and to implementing emerging best practice in a timely manner.
12. Updates to This Policy
This policy will be reviewed at least annually and updated whenever we introduce a new AI system or make a significant change to an existing one. Where changes are material, we will notify users by email and via a platform notice at least 14 days before the changes take effect.
The date at the top of this document reflects when it was last updated. Continued use of the Platform after an update takes effect constitutes acceptance of the revised policy.
13. Contact
If you have questions about this policy, want to raise a concern about how our AI systems have affected you, or wish to exercise any of your rights, please send an email to
