Last updated: 20/03/2026
1. Our approach
Diversit Limited provides data and artificial intelligence consultancy services. We believe AI should be designed, assessed, deployed and governed in a way that is lawful, human-centred, proportionate and accountable.
This page summarises our public-facing commitments. It does not replace project-specific contracts, compliance assessments, statements of work, data protection documentation, sector-specific controls or legal advice.
2. What we mean by responsible AI
In our work, we aim to support AI practices that are:
- lawful and aligned with applicable regulation;
- transparent and appropriately documented;
- subject to human oversight where needed;
- fair and mindful of bias and discriminatory impacts;
- secure and resilient;
- privacy-aware and data-governed;
- proportionate to context, use case and risk.
3. Our commitments
Where relevant to the services we provide, we seek to:
Support lawful use
Design and advise on AI-related work in a way that considers applicable legal and regulatory requirements, including data protection, confidentiality, intellectual property, sector-specific rules and emerging AI regulation.
Promote risk-based governance
Tailor governance recommendations to the nature of the use case, data involved, users affected and the materiality of the outcome.
Encourage human oversight
Recommend clear human accountability, escalation and review paths, especially where outputs may affect people, regulated decisions, operational control or material business outcomes.
Address transparency
Encourage clear documentation of the purpose, boundaries, assumptions, limitations and intended use of AI systems and outputs.
Consider data quality and provenance
Promote appropriate controls around input data quality, access, retention, confidentiality and lawful sourcing.
Consider fairness and impact
Where relevant, consider bias, error patterns, affected groups and foreseeable misuse scenarios as part of an appropriate governance process.
Promote security and resilience
Support security-aware design, access controls, testing, monitoring and incident response planning proportionate to the solution.
Build capability and literacy
Support AI literacy, governance awareness and practical understanding among relevant stakeholders.
4. Prohibited, unlawful and high-risk uses
Diversit Limited does not intend its website, materials or standard advisory content to support prohibited or unlawful uses of AI.
Where an engagement may involve sensitive, regulated, safety-critical, employment-related, education-related, healthcare-related, insurance-related, financial or public-sector contexts, additional controls may be needed. Those controls are assessed case by case and should be documented in the relevant project terms and governance materials.
5. AI outputs and decision-making
Unless expressly agreed in writing:
- Diversit Limited does not warrant that AI-generated or AI-assisted outputs will be error-free, complete, unbiased or suitable without human review;
- AI outputs should not be treated as a substitute for legal, regulatory, financial, medical or other specialist advice;
- final responsibility for deployment choices, acceptance criteria, operational controls and decision-making remains with the client or end user.
6. Client-specific compliance
Compliance obligations under the EU AI Act and related laws depend on the facts of each engagement, including the role of the parties, the type of system, the sector, the intended use, the geography and the contractual allocation of responsibilities.
For that reason:
- this page is a general statement of approach only;
- detailed compliance responsibilities are defined in project-specific agreements and governance materials;
- clients remain responsible for obtaining their own legal advice where needed.
7. Continuous review
AI regulation and guidance continue to develop. We review our policies and advisory approach periodically and may update this page as the legal and technical environment evolves.
8. Contact
Questions about this policy can be sent to:
privacy@diversit.io
