Our commitment
AI can create real value, and it can also cause real harm when it is built or used carelessly. UFUQORA AI is committed to developing and operating AI services that are safe, fair, transparent and accountable, and to helping our customers use AI the same way.
Principles
- Human-centred. AI should support people and their decisions, not replace accountability for them.
- Transparent. People should know when they are interacting with AI or viewing AI-generated content.
- Fair. We work to identify and reduce unfair bias in the systems we provide.
- Safe and secure. We test our services, protect them from misuse and monitor them in operation.
- Privacy-respecting. We minimise personal data and protect the data we are trusted with.
- Accountable. We assign clear ownership for our AI services and the decisions we make about them.
Frameworks we align with
We design our practices to align with recognised frameworks, including:
- Singapore: the Model AI Governance Framework published by IMDA and the PDPC, including the Model AI Governance Framework for Generative AI, the AI Verify testing framework, and the Personal Data Protection Act 2012;
- United States: the NIST AI Risk Management Framework; and
- European Union: the EU AI Act (Regulation (EU) 2024/1689), including its transparency obligations for AI systems and AI-generated content, and the General Data Protection Regulation.
Alignment with a framework describes how we have designed our practices; it is not a certification by the body that published it.
Governance across the lifecycle
- Before deployment, we assess the intended use, the people it may affect, data sources and foreseeable risks, and we decline uses that conflict with our Acceptable Use Policy.
- During development and configuration, we evaluate models for quality, robustness and relevant bias, using domain experts where the use case requires it.
- In operation, we monitor performance and incidents, keep records of significant changes and act on feedback.
Human oversight
For uses that may significantly affect people, such as in finance, healthcare, legal or employment contexts, our services are designed to support human review rather than make autonomous decisions. We help customers put appropriate oversight in place.
Transparency and labelling
AI-generated video and other synthetic media produced with our services carry labels and, where technically feasible, machine-readable markings that identify them as AI-generated. We tell people when they are interacting with an AI system where that is not obvious.
Customer data and model training
We do not use customer prompts, outputs or uploaded data to train models unless the customer has agreed to it. Fine-tuning on customer data takes place in environments dedicated to that customer.
Third-party models
When we provide models developed by others, we review their documentation, licences and known limitations before offering them, and we pass relevant information on to our customers.
Feedback
We welcome questions and concerns about how we use AI. Please email [email protected] with “Responsible AI” in the subject line.