UNESCO Draft Recommendation on the Ethics of Artificial Intelligence:

A Crucial Milestone for an International Framework

By: Romina Ruszin

For billions of people, Artificial Intelligence (hereinafter, “AI”) has become an integral part of everyday life. It recommends content on streaming services, facilitates online banking, and plans the fastest route to a destination.[1] The cognitive technology behind these mundane tasks is the same one that could provide support for millions in completing secondary school and create 3.3 million jobs.[2] It is also the one whose malicious engineering and implementation poses a threat to privacy and dignity, enables mass surveillance, and has an adverse effect on the environment.[3] Currently, no global instruments exists to reap the benefits and address the risks of AI.[4] The UNESCO Draft Recommendation on the Ethics of Artificial Intelligence,[5] adopted on November 23, 2021, aspires to rectify this dearth of the international community.[6]

The Recommendation’s Content

The Recommendation aims to establish a standard-setting legal infrastructure.[7] It strives to insert a human rights-based moral ground into AI technologies’ manufacturing and employment to ensure that AI enhances, rather than deteriorates quality of life.[8] The Draft intends to establish AI as a tool to support humanity by proposing measures to control current and anticipated issues alike.[9]

Data Protection

The Recommendation strongly pushes for the respect and promotion of privacy and data protection.[10] It proposes that individuals are guaranteed agency over their personal data.[11] In fact, The Draft goes as far as to suggest that people should be able to readily access and erase such information stored by AI systems,[12] except for exceptional circumstances outlined by law.[13] The Recommendation encourages a multi-pronged approach for the accomplishment of this goal. For one, educational, private, and governmental institutions should aim to increase individuals’ cognizance of AI systems and their rights associated with its usage.[14] On another note, AI actors should conduct Privacy Impact Assessments to evaluate the societal and ethical effects of their creations.[15] Thanks to such precautionary measures, it would be guaranteed that personal data is protected throughout the life cycle of AI systems: their design, use, and destruction.[16] At the heart of any action taken must be conscientious Member States, who establish and reinforce such policies in order to insure the security of personal data.[17]

Social Scoring and Mass Surveillance

The Recommendation explicitly asserts that AI technology should not be used for social scoring or mass surveillance purposes.[18] The use of AI should always be appropriate and proportional for a legitimate aim, not infringe on human rights, and be based on rigorous scientific foundations.[19] Neither social scoring, the ranking of individuals based on reductive metrics,[20] nor mass surveillance, the collection, storing, and the analysis of private communications satisfy the requirements laid out by the Recommendation.[21] Due to the invasive nature of AI in general, the Draft emphasizes the importance of human supervision over decisions made by AI.[22] It is therefore paramount that regulatory frameworks assign responsibility and accountability to humans and bar AI from obtaining legal personality.[23]

Environment

The Recommendation stresses the importance of AI methods that are data, energy, and resource-efficient.[24] This is not only crucial to reducing AI system’s environmental impact and ensuring its compliance with environmental policies,[25]but also to realizing AI’s potential as a tool to tackle environmental issues.[26]Accordingly, the Draft requests that governments conduct Environmental Impact Assessments that take into account AI technologies’ carbon footprint, energy consumption, and the impact of raw material extraction required to assemble and operate AI systems.[27] It further asks Member States to develop benchmarks that ensure that the application of AI will have the intended effect.[28] These safeguards would also justify its use under the principle of proportionality.[29] The Recommendation is strict to point out that if this is impossible, the precautionary principle should be applied; when Environmental Impact Assessments suggest a disproportionately negative impact on environments, the use of AI should be prohibited.[30]

Implementation

The Draft recognizes the ambitious nature of its goals and the inherent difficulties signatories will face in the bid to achieve them. It, therefore, proposes mechanisms to enable the effective implementation of the Recommendation.[31] It encourages Member States and private actors to create positions for AI Ethics Officers who would oversee auditing and monitoring efforts.[32] The establishment of an independent network for such Officers is further suggested to promote the initiative.[33] Moreover, Readiness Assessment Technology will determine and help enhance Member States’ infrastructural and legal capacity on their trajectory to ethical implementation of AI.[34] The Recommendation also suggests the development of a UNESCO methodology for Ethical Impact Assessment,[35] which strives to assess the foreseeable impact of AI technology on individuals, societies, and environments.[36] The program also entails government officials, policymakers, and AI engineers’ training in EIA methodology since these players’ knowledge and morality will be the very foundation on which the ethical implementation of AI stands on.[37]

Ultimately, it will be up to individual governments’ discretion and vigilance whether the Recommendation is translated into national legislation and policy.[38] The Recommendation, and a draft at that, is a non-binding piece of soft law.[39]For now, it only establishes a political and moral obligation to respect its foundation.[40] Nevertheless, the Recommendation is a paramount milestone on the path to adopting a binding international instrument one day. Until then, the ball is in Member States’ court.

[1]‘Artificial Intelligence in Daily Life with Examples’ (Medium, 11 August 2021) <https://becominghuman.ai/artificial-intelligence-in-daily-life-with-examples-a363502086ff> accessed 25 January 2022. [2] UNESCO ‘Recommendation on the Ethics of Artificial Intelligence’ (UNESCO, 6 December 2021) <https://en.unesco.org/artificial-intelligence/ethics> accessed 25 January 2022. [3] ‘UNESCO Member States Adopt the First ever Global Agreement on the Ethics of Artificial Intelligence’ (UNESCO, 25 November 2021) <https://en.unesco.org/news/unesco-member-states-adopt-first-ever-global-agreement-ethics-artificial-intelligence> accessed 25 January 2022. [4] UNESCO (n 2). [5] UNESCO ‘Draft Text of the Recommendation on the Ethics of Artificial Intelligence’ (25 June 2021) SHS/IGM-AIETHICS/2021/JUN/3 Rev 2. [6] Eugenio Vargas Garcia ‘UNESCO’s Recommendation on the Ethics of AI: Why it Matters And What to Expect From it’ (The Good AI, 24 November 2021) <https://thegoodai.co/2021/11/24/unescos-recommendation-on-the-ethics-of-ai-why-it-matters-and-what-to-expect-from-it/> accessed 25 January 2022. [7] UNESCO ‘Draft text of the Recommendation on the Ethics of Artificial Intelligence (n 5). [8] UNESCO (n 2). [9] ibid. [10] UNESCO ‘Draft text of the Recommendation on the Ethics of Artificial Intelligence (n 5) art 32. [11] ibid. [12] ibid art 73. [13] ibid. [14] ibid. [15] ibid art 34. [16] ibid. [17] ibid art 74. [18] ibid art 26. [19] ibid. [20] Amnesty International, ‘EUROPE: Proposed Legislation Too Weak to Protect Us From Dangerous AI Systems’ (Amnesty International, 21 April 2021) <https://www.amnesty.org/en/latest/news/2021/04/eu-legislation-to-ban-dangerous-ai-may-not-stop-law-enforcement-abuse/> accessed 25 January 2022. [21] ibid. [22] UNESCO ‘Draft text of the Recommendation on the Ethics of Artificial Intelligence (n 5) art 68. [23] ibid [24] ibid art 86. [25] ibid art 84. [26] ibid art 85. [27] ibid art 86. [28] ibid. [29] ibid. [30] ibid. [31] ibid art 49. [32] ibid art 58. [33] ibid. [34] ibid art 131(b). [35] ibid art 49. [36] ibid art 5. [37] ibid art 4. [38] Garcia (n 6). [39] ibid. [40] ‘UNESCO instruments’ (Right to Education Initiative) <https://www.right-to-education.org/page/unesco-instruments> accessed 25 January 2022. [photograph of] Artificial Intelligence (Forbes, 8 May 2020) <https://imageio.forbes.com/specials-images/imageserve/5eb49cca1b61700007592462/5-Reasons-Why-Artificial-Intelligence-Really-Is-Going-To-Change-Our-World/960x0.jpg?fit=bounds&format=jpg&width=960> accessed 18 March 2022.