What organisational certification is for.
Ethical AI certification is intended to turn commitments into evidence. It will help an organisation demonstrate that responsibility is reflected in governance, roles, controls, records, decision-making and routes for people to challenge an outcome.
The assessment will focus on how AI is used in real situations, not only on the presence of policies. Its purpose will be to make strengths visible, identify gaps and create a practical basis for improvement.
The question is not only whether an organisation has an AI policy. It is whether people can understand, challenge and change how AI affects them.
What the assessment will examine.
The certification pathway is being developed around the six assessment axes of the PACT Standard. Together they connect organisational practice with the experience of the person affected by technology.
- Human oversight
- Whether responsible people have the authority, information, competence and time to intervene effectively.
- Transparency
- Whether the organisation can explain where AI is used, what role it plays and what its relevant limitations are.
- Privacy
- Whether personal data, inference and profiling are limited, justified and governed throughout the lifecycle.
- Safety
- Whether foreseeable failures are tested, monitored and met with effective fallback and incident response.
- Non-discrimination
- Whether unequal outcomes are examined across affected groups and corrected when they appear.
- Social impact
- Whether the organisation considers effects on work, relationships, participation, access and communities.
Who it is intended for.
The future process is being designed for organisations that make consequential choices about AI, including:
- public institutions and local government,
- technology companies and product teams,
- schools, universities and education providers,
- social organisations and essential service providers,
- organisations procuring AI-enabled systems from suppliers.
Final eligibility criteria and certification scopes will be confirmed when the programme is launched.
How the process is expected to work.
The method is still being prepared. The current direction is a staged process that connects documentary evidence with the way people actually work.
- Scope and readiness. Define the organisation, services, AI-enabled processes and affected groups included in the assessment.
- Evidence. Gather policies, responsibilities, risk records, tests, monitoring results, complaints and examples of intervention.
- Assessment. Review the evidence and operating practice against the six PACT axes.
- Improvement. Resolve material gaps and verify that corrective actions work in practice.
- Decision and continued review. Reach a bounded decision for a defined scope and period, with expectations for monitoring and reassessment.
What certification will not mean.
- It will not mean that an organisation or AI system has zero risk.
- It will not be a blanket approval of every technology in use.
- It will not replace legal, regulatory or sector-specific duties.
- It will not be a one-time substitute for ongoing oversight.