Purpose
AI must benefit people, respect their rights, and operate within strong ethical boundaries. Humans remain accountable for every outcome generated by AI at all times.
Core Commitments
We commit to protecting autonomy, privacy, fairness, and human dignity through the following behaviors:
- Collect only the minimum necessary data.
- Avoid secretly inferring sensitive attributes.
- Enforce lawful, documented sourcing for training data.
- Treat fairness as measurable, monitored, and traceable.
Responsible Design and Deployment
AI must be explainable in proportion to risk, and design artefacts must capture context, limitations, and fairness targets.
- Document purpose and intended users.
- Identify known limitations and underlying assumptions.
- Capture dataset lineage and quality checks.
- Define fairness targets and monitor continuously.
- Act immediately when real-world harm or unfair patterns emerge.
Security and Integrity
We secure model artefacts, training data, and inference traffic against adversaries.
- Apply strong authentication and least-privilege access.
- Log relevant events and preserve audit trails.
- Escalate incidents based on severity, with critical events reported to leadership within 24 hours.
Governance and Oversight
All AI systems must be registered in the central AI registry and monitored after deployment.
- High-impact AI requires approvals from Legal, AI Governance, Security, and Privacy before production release.
- Models are retired or replaced when no longer appropriate or safe.
Usage Boundaries
The company prohibits the following uses of AI and expects vendors to uphold equivalent standards.
- Social scoring or ranking of individuals.
- Biometric categorization based on sensitive traits.
- Psychological manipulation or coercion.
- Surveillance that infringes on rights or freedoms.
People and Accountability
We invest in AI literacy, expect every employee to report misuse or harm, and review this charter annually to incorporate new regulations, societal impact, and risk learnings.