Loading
Responsible AI is the umbrella discipline for designing, developing, deploying, and operating AI systems in ways that are ethical, fair, transparent, accountable, safe, and respectful of privacy. It sits above AI ethics (the values) and AI governance (the management framework): ethics supplies the principles, governance supplies the structures, and responsible AI practices, fairness testing, explainability, monitoring, translate both into operational reality. Every major technology company and several governments now publish their own responsible AI framework, and while the specific principles vary, the underlying themes converge.
Run the free AI Health CheckResponsible AI, the discipline of designing, developing, deploying, and using AI in ways that align with stated values: typically fairness, accountability, transparency, safety, privacy, and human autonomy.
Responsible AI is the umbrella discipline that sits over AI governance, AI ethics, and AI safety. Microsoft, Google, Anthropic, OpenAI, IBM, and most major enterprises maintain Responsible AI teams and published principles. The principles are similar across organisations: the substance is in how the principles are operationalised through review processes, model cards, audit, red teaming, and disclosure.
Source: OECD AI Principles; ISO/IEC 42001
Responsible AI is not one document, it is a family of published frameworks from technology companies, intergovernmental bodies, and national governments. None of these four are legally binding on their own, they are voluntary commitments, though many organisations now map them to binding law such as the EU AI Act where it applies.
21 June 2022 · Internal corporate requirement
Fairness, Reliability and Safety, Privacy and Security, Inclusiveness, Transparency, Accountability (6 principles)
Adopted 22 May 2019, updated 3 May 2024 · Non-binding Council Recommendation, 47 adherents incl. the EU
Inclusive growth and well-being; human rights, democratic values and fairness; transparency and explainability; robustness, security and safety; accountability (5 principles)
7 November 2019 · Voluntary (Dept of Industry, Science and Resources)
Human, social and environmental wellbeing; human-centred values; fairness; privacy protection and security; reliability and safety; transparency and explainability; contestability; accountability (8 principles)
First launched Jan 2019, 2nd Edition 21 Jan 2020 · Voluntary (PDPC/IMDA)
Two guiding principles (explainable/transparent/fair decisions; human-centric systems) across four practice areas: internal governance, human involvement, operations management, stakeholder communication
What is responsible AI?
Responsible AI is an umbrella term for the principles, standards, and governance practices organisations and governments use to ensure AI systems are designed, developed, and deployed safely, fairly, transparently, and with accountability for their impact on people and society. It is typically operationalised through published principle sets backed by internal processes such as impact assessments and oversight review.
Is responsible AI the same thing as the EU AI Act?
No. The EU AI Act is binding legislation with legal penalties. Responsible AI more broadly refers to voluntary principles and internal standards, such as Microsoft's Responsible AI Standard, the OECD AI Principles, Australia's AI Ethics Principles, and Singapore's Model AI Governance Framework, that predate and are independent of the EU AI Act, though many organisations now map these frameworks to binding law where it applies.
Who publishes responsible AI frameworks?
Three broad groups: individual technology companies publishing their own internal standards (Microsoft's Responsible AI Standard, Google's AI Principles); intergovernmental bodies (the OECD's AI Principles, UNESCO's Recommendation on the Ethics of AI); and national or regional governments (Australia's AI Ethics Principles, Singapore's Model AI Governance Framework).
Are responsible AI frameworks legally binding?
Generally no. Microsoft's Standard is an internal corporate requirement, not law. The OECD AI Principles are a Council Recommendation, a non-binding legal instrument adhered to by 47 countries plus the EU. Australia's AI Ethics Principles are explicitly voluntary. Singapore's Model AI Governance Framework is expressly non-binding guidance. This contrasts with the EU AI Act, which is binding, enforceable law.
What common themes appear across major responsible AI frameworks?
Despite different authors, most frameworks converge on the same core ideas: fairness and non-discrimination, transparency and explainability, privacy and security, reliability and safety, human oversight or human-centred values, and accountability. Microsoft's six principles, the OECD's five, Australia's eight, and Singapore's framework all include variants of these themes.
How often are responsible AI frameworks updated?
Periodically, as the technology changes. The OECD AI Principles were updated on 3 May 2024, five years after their 2019 adoption, specifically to address generative and general-purpose AI. Singapore issued a Second Edition of its Model AI Governance Framework about a year after the first, and later published a related framework specifically for generative AI. Microsoft's Standard is versioned, its current public version dates to June 2022.