Appendix 2: Comparison of Ethical AI Principles

	A	C	D	E	F	G	H	I	J	K	L
1	Global Principles			Country/Region Principles			Meta-analyses of multiple principles			W3C Principles (WIP)
2	UNESCO Recommendation on the ethics of artificial intelligence	EU High Level Expert Group on AI: Ethics Guidelines for Trustworthy AI Principles (P) and Requirements	OECD AI Principles / G20 AI Principles (the two are the same)	China: Governance Principles for the New Generation Artificial Intelligence	Digital Dubai AI Principles	Japan: Social Principles of Human-centric AI	Principled Artificial Intelligence: Mapping Consensus in Ethical and Rights-based Approaches to Principles for AI by Harvard Berkman Klein Centre	Associated Principles	Hagendorf: The Ethics of AI Ethics: An Evaluation of Guidelines (Including the top 10 principles)	W3C TAG Ethical Web Principles
3	VALUES
4	Respect, protection and promotion of human rights and fundamental freedoms and human dignity			Harmony and human friendly: The goal of AI development should be to promote the well-being of humankind. It should conform to human values and ethical principles, promote human-machine harmony, and serve the progress of human civilization. The development of AI should be based on the premise of ensuring social security and respecting human rights, and should prevent misuse, abuse and evil use of AI technology by all means.	We will give AI systems human values and make them beneficial to society We will promote human values, freedom and dignity	Dignity Human-centric	PROMOTION OF HUMAN VALUES	Improve social benefits, Values and human development and Access to technology	common good, sustainability, well-being	The web must enable freedom of expression	We will create web technologies and platforms that encourage free expression, where that does not contravene other human rights. Our work should not enable state censorship, surveillance or other practices that seek to limit this freedom. This principle must be balanced with respect for other human rights, and does not imply that individual services on the web must therefore support all speech. (For example: hate speech, harassment or abuse may reasonably be denied a platform).
5	Environment and ecosystem flourishing			Inclusion and sharing: promote green development to meet the requirements of environmental friendliness and resource conservation		Sustainability			common good, sustainability, well-being
6	Ensuring diversity and inclusiveness		Inclusive growth, sustainable development and wellbeing	Inclusion and sharing: support to the vulnerable groups to adapt, and efforts to eliminate digital divide	We will share the benefits of AI throughout society	Diversity & inclusion			solidarity, inclusion, social cohesion
7	Living in peaceful, just and interconnected societies									There is one web	When we are adding new web technologies and platforms, we will build them to cross regional and national boundaries. People in one location should be able to view web pages from anywhere that is connected to the web.
8
9	PRINCIPLES
10	Proportionality and Do No Harm	(P) Prevention of harm			AI systems should not be able to autonomously hurt, destroy or deceive humans We will give AI systems human values and make them beneficial to society					The web should not cause harm to society The web must support healthy community and debate	When we are adding a feature or technology to the web, we will consider what harm it could do to society or groups, especially to vulnerable people. We will prioritize potential benefits for web users over potential benefits to web developers, content providers, user agents, advertisers or others in the ecosystem, in line with the priority of constituencies. We will ensure the requirements and views of marginalized communities and underrepresented groups are heard and respected. We are building technologies and platforms for distributing ideas, for virtual interaction, and for mass collaboration on any topic. While those tools can be used for good, they can also be used for spreading misinformation, revealing private personal information (doxing), harassment, and persecution. We will consider these risks in the work we do, and will build web technologies and platforms that respect individuals' rights and provide features to empower them against dangers like these.
11	Safety and security	Technical robustness and safety	Robustness, security and safety	Safety and Controllability: The transparency, interpretability, reliability, and controllability of AI systems should be improved continuously to make the systems more traceable, trustworthy, and easier to audit and monitor. AI safety at different levels of the systems should be ensured, AI robustness and anti-interference performance should be improved, and AI safety assessment and control capacities should be developed.	AI systems will be safe, secure and controllable by humans	Security	PROTECTION AND SECURITY	Protection, Security and trust, Predictability and Security of the design	safety, cybersecurity	Security and privacy are essential	When we add features to the web platform, we are making decisions that may change the ability of people to protect their personal data. This data includes their conversations, their financial transactions and how they live their lives. We will start by creating web technologies that create as few risks as possible, and will make sure people understand what risks they are taking when they use the web.
12	Fairness and non-discrimination	(P) Fairness Diversity, non-discrimination and fairness	Human-centred values & fairness	Fairness and Justice: The development of AI should promote fairness and justice, protect the rights and interests of all stakeholders, and promote equal opportunities. Through technology advancement and management improvement, prejudices and discriminations should be eliminated as much as possible	We will make AI systems fair	Fairness, Accountability, and Transparency	JUSTICE AND NO DISCRIMINATION	Non-discrimination and prevention of bias, Justice Inclusion in design, Inclusion in impact, Date representative and of quality and Equality	fairness, non-discrimination, justice	The web is for all people	Anyone should be able to meaningfully participate in the creation of specifications, user agents, and content, and the platform should enable a fully accessible end user experience. We will build internationalization and localization capabilities into our specifications and websites. We will accommodate people on low bandwidth networks and with low specification equipment. The web platform and the tools we use to create it must be accessible to people with disabilities, including visual, auditory, physical, speech, cognitive, language, learning, and neurological disabilities.
13	Sustainability	Societal and environmental wellbeing	Inclusive growth, sustainable development and wellbeing	Inclusion and sharing: promote green development to meet the requirements of environmental friendliness and resource conservation					common good, sustainability, well-being	The web must be an environmentally sustainable platform	This includes maximizing the lifespan of physical devices through maintaining compatibility, as well as minimizing data storage and processing requirements.
14	Right to Privacy, and Data Protection	Privacy and data governance		Respect for Privacy: AI development should respect and protect the privacy of individuals and fully protect an individual’s rights to know and to choose.	We will respect people’s privacy	Privacy	PRIVACY	Privacy, Control over the use of data, Consent, Privacy by design, Recommendations for data protection law, Ability to restrict processing, Right to rectification and Right to delete unwanted data.	privacy protection	Security and privacy are essential	When we add features to the web platform, we are making decisions that may change the ability of people to protect their personal data. This data includes their conversations, their financial transactions and how they live their lives. We will start by creating web technologies that create as few risks as possible, and will make sure people understand what risks they are taking when they use the web.
15	Human oversight and determination	Human agency and oversight		Safety and Controllability: The transparency, interpretability, reliability, and controllability of AI systems should be improved continuously	AI systems will be safe, secure and controllable by humans		HUMAN CONTROL OF TECHNOLOGY	Human control of technology, Human review of automated decisions, and Ability to opt out of automated decisions	human oversight, control, auditing
16	Transparency and explainability	(P) Explicability Transparency	Transparency and explainability	Respect for Privacy: AI development should respect and protect the privacy of individuals and fully protect an individual’s rights to know and to choose. Safety and Controllability: The transparency, interpretability, reliability, and controllability of AI systems should be improved continuously	We will make AI systems as explainable as technically possible We will make AI systems transparent	Fairness, Accountability, and Transparency	TRANSPARENCY AND EXPLANABILITY	Explainability, Transparency, Open source data and algorithms, Notification when interacting with AI, Notification when AI makes a decision about an individual, Regular information and reporting requirements, Right to be informed, and Open Contracting (for governments)	transparency, openness explainability, interpretabiliy	The web must make it possible for people to verify the information they see The web is transparent	We have a responsibility to build web technologies to counter misinformation, allowing information sources to be traceable and facts to be checkable. The concept of origin and source is core to the web's security model. We will make sure the new web technologies we create do not work against this architectural principle. We will always make sure it is possible to determine how a web application was built and how the code works. Furthermore, we will always make sure it is possible to audit and inspect web applications and underlying software for security, privacy or other considerations.
17	Responsibility and accountability	Accountability	Accountability	Shared Responsibility: AI developers, users and other related stakeholders should have a high sense of social responsibility and self-discipline, and should strictly abide by laws, regulations, ethical principles, technical standards and social norms. AI accountability mechanisms should be established to clarify the responsibilities of researchers, developers, users, and relevant parties. Users of AI products and services and other stakeholders should be informed of the potential risks and impacts in advance. Using AI for illegal activities should be strictly prohibited.	We will make AI systems accountable	Fairness, Accountability, and Transparency	ACCOUNTABILITY	Recommendation for new regulations, Impact assessment, Assessment and audit requirements, Verifiability and replicability, Legal responsibility, Appeal capacity, Environmental responsibility, Creation of a monitoring body.	accountability
18	Awareness and literacy					Education / AI Literacy
19	Multi-stakeholder and adaptive governance and collaboration			Agile Governance: The governance of AI should respect the underlying principles of AI development. In promoting the innovative and healthy development of AI, high vigilance should be maintained in order to detect and resolve possible problems in a timely manner. The governance of AI should be adaptive and inclusive, constantly upgrading the intelligence level of the technologies, optimizing management mechanisms, and engaging with muti-stakeholders to improve the governance institutions. The governance principles should be promoted throughout the entire lifecycle of AI products and services. Continuous research and foresight for the potential risks of higher level of AI in the future are required to ensure that AI will always be beneficial for human society. Open and collaboration: With full respect for the principles and practices of AI development in various countries, international dialogues and cooperation should be encouraged to promote the formation of an international AI governance framework with broad consensus. Cross-disciplinary and cross-boundary exchanges and cooperation should be encouraged in the development of AI. Coordinated interactions should be fostered among international organizations, government agencies, research institutions, educational institutions, industries, social organizations and the general public in the development and governance of AI.	We will plan for a future in which AI systems become increasingly intelligent We will govern AI as a global effort	Innovation: through a Global perspective we must promote diversification and cooperation between industry-academia-public-private sectors, through the development of human capabilities and technology.	PROFESSIONAL RESPONSIBILITY	Multi-stakeholder collaboration, Responsible design, Consideration of long-term effects, Scientific integrity and Accuracy and precision	Science-policy link		We will build new web technologies in a collaborative manner according to open processes (for example, the W3C process), and adhering to codes of conduct (such as the W3C Code of Ethics and Professional Conduct).
20
21	OTHERS NOT IN UNESCO
22		(P) Respect for human autonomy		Inclusion and Sharing: AI should promote coordinated development by promoting the transformation and upgrading of all industries, and by narrowing regional disparities; AI should promote inclusive development through better education and training, support to the vulnerable groups to adapt, and efforts to eliminate digital divide; AI should promote shared development by avoiding data and platform monopolies, and promoting open and fair competition.		Fair competition				The web must enhance individuals' control and power	We recognize that web technologies can be used to manipulate and deceive people, complicate isolation, and encourage addictive behaviors. We seek to mitigate against these potential abuses and patterns when creating new technologies and platforms, and avoid introducing technologies that increase the chance of people being harmed in this way. As part of this, we favor a decentralized web architecture that minimizes single points of failure and single points of control. We will also build web technologies for individual developers as well for developers at large companies and organizations. The web should enable do-it-yourself developers.
23						Innovation				NOT RELEVANT FOR ETHICAL PRINCIPLES?
24										The web is multi-browser, multi-OS and multi-device	We will not create web technologies that encourage the creation of websites that work only in one browser, or only on particular hardware. We expect that content provided by accessing a URL should yield a thematically consistent experience when someone is accessing it from different devices. The existence of multiple interoperable implementations enables competition, and thus a variety of choices for web users.
25										People should be able to render web content as they want	People must be able to change web pages according to their needs. For example, people should be able to install style sheets, assistive browser extensions, and blockers of unwanted content or scripts or auto-played videos. We will build features and write specifications that respect peoples' agency, and will create user agents to represent those preferences on the web user's behalf.
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100