We build AI systems for our clients and use AI in our work. We believe that AI generates positive value only where the standards of ethics and responsibility governing it are explicitly defined. We propose the following guiding principles and encourage our clients and partners to commit to them for responsible use of AI.
| 1. Accountability | Ensure clear accountability for every AI system built and in use, providing meaningful human review in the loop. |
|---|---|
| 2. Data security and governance | Build AI systems that consider privacy, security, confidentiality, and intellectual property rights with due attention to data usage. |
| 3. Accuracy and reliability | Measure AI system performance against agreed criteria using trustworthy tools and methodologies. |
| 4. Fairness | Ensure human oversight of AI systems from the perspectives of diversity, equity, and inclusion, mitigating unfair discrimination and bias risks. |
| 5. Understandability and documentation | Develop AI systems whose operations and decision-making processes related to the outputs can be understood by people. Document the key design decisions, system protocols, and data structures, to make the development traceable and explainable. |
| 6. Technology due diligence | Assess third-party models, APIs, datasets, and AI vendors before selection and adoption. |
| 7. Social impact | Ensure that AI systems contribute to social well-being and environmental sustainability, protecting human life, health, and property at every stage of development and use. |
| 8. Maintenance responsibility | Conduct continual evaluation and monitoring of AI systems, taking action where necessary to ensure adherence to the principles. |