By Emile Loza de Siles*
The availability and uses of artificial intelligence, including machine learning, deep learning, and other data and algorithmic systems, and robotic systems enabled by artificial intelligence (collectively, “artificial intelligence” or “AI”)[1] are exploding. Beyond doubt, some AI innovations are profoundly important and useful.[2] Some unlock new frontiers far beyond the effective reaches of humans on this blue planet.[3] Some AI innovations are unfolding here on Earth, transforming our lives, although not without safety concerns.[4]
Apart from a few AI applications,[5] however, artificial intelligence domains are largely lawless.[6] For example, humanoid childlike robotic systems are marketed for use by human children, capturing the children’s every move and utterance with vision systems and sensors masked behind large doe-like “eyes” and false ears of the systems.[7] Predictive policing systems target predominantly black and brown neighborhoods.[8] The federal government deployed facial recognition systems at airports in the United States without the consent of the governed or the traveling uber-surveilled whose faces were ingested and whose biometric data now likely permanently persist in some unknown server farm and are farmed out for other unknown uses. An autonomous vessel patrols the sea, promising greater protection of now-absent personnel from weather, submarine attacks, and pirates,[9] but easily conceivable as a platform for potentially rogue autonomous weapon systems.[10]
State and local governments are integrating artificial intelligence (“AI’), including facial recognition technology, into their operations at an increasingly rapid pace, as are national governments.[11] Globally, government spending on AI represented almost one-fifth of the world’s AI market in 2019.[12] Moreover, private vendors of AI systems are quickly becoming embedded into government functions, which, consequently, are increasingly obscured and beyond any meaningful rule of law.[13]
As a result, the Data Industrial Complex is real and has significant and potentially devastating impacts upon people, their families and communities, legitimate governments subject to the meaningful rule of law, and democracy itself.[14] Numerous instances of major errors and unfair biases in AI systems and of AI-mediated bias and discrimination have been widely reported in the popular, scientific, and legal literature.[15] Despite the drumbeating reports of such algorithmic abuses,[16] the vacuum of AI governance persists, and people suffer as a result and at a potentially exponentially increasing scale.[17]
I. A Multi-Pronged Legal Development Strategy to Institute Rapid and Informed Artificial Intelligence Governance
This Paper offers actionable clarity by mapping out a multipronged legal development strategy by which AI governance can be established more efficiently, rapidly, and with greater technological competency and foresight than traditional and linear legal development approaches. It first presents technical standards and best practices as a “soft law” legal development strategy by which to competently examines artificial intelligence systems and uses and the associated legal and ethical concerns. Second, it argues for a comprehensive AI-rationalization initiative by which existing civil rights law should be interpreted and thereby reach artificial intelligence domains that currently are largely lawless. Third, it casts a view toward new AI-specific governance to fill any gaps left unaddressed by the AI soft law and AI-rationalization prongs of the proposed legal strategy.
A. Technical Standards as Soft Law
As the first prong of the proposed AI legal development strategy, this work looks to leverage the extensive work toward trustworthy AI spurred by numerous ethical technology movements.[18] International multi-stakeholder efforts are underway to establish AI technical standards, best practices, and ethical guidelines (collectively, “AI standards”).[19] These AI standards initiatives, however, go beyond the “merely technical” to focus on legal compliance as a bare minimum standard and human and planetary well-being are the ultimate aims of these standards.[20]
Soft law development principles from the international legal domain[21] can inform and expedite competent AI governance. AI accuracy, auditability, explainability, fairness, non-bias, non-discrimination, human-centeredness, and other substantive and procedural protections can flow from this growing body of AI soft law. Married with private ordering, these protections can be made enforceable as contractually enforceable matters of technical specification and market-demanded product feature.[22] Although private contractual enforceability, particularly among tech giants such as Microsoft, Apple, Amazon, Google, and others, may not necessarily lead to government-enforceable AI law, that enforceability nevertheless brings important power to motivate vendors and, as customers, governments toward trustworthy AI.[23] If the pattern observed in cybersecurity and other fields of technological innovation holds true, AI standards may be rapidly integrated within existing legal regimes that are enforceable by governments and private litigants, however.[24]
B. A Governance Initiative to AI-Rationalize Existing Law
As the second prong of the proposed AI legal development strategy, existing laws should undergo a process of “AI rationalization” to make their interpretation and application within AI contexts clearer for bench, bar, industry, and individuals. A broad legislative and regulatory initiative is needed. Given the risks to people from government uses of facial recognition and other predictive AI systems,[25] however, consumer protection and civil rights laws should be prioritized.
Although legislatures have discussed aspects of artificial intelligence and some AI bias and discrimination laws have been proposed,[26] they have effectuated little actual AI governance and no AI rationalization of existing laws.[27] On its enormous and enormously powerful part, the federal administrative state is largely engaged in a dangerous game of delay, kicking the AI governance can indefinitely down the road in hopes that others will fill the void.[28]
The U.S. Federal Trade Commission (“FTC”)[29] and a few other agencies[30] have begun the process of AI-rationalizing the laws within their jurisdictional scopes, however. The FTC, for one, has repeatedly demonstrated the success of such interpretative and contextualizing legal development approaches to technological innovations.[31] Following such a model will ensure that existing civil rights and other protective laws and doctrines are brought onto new and essential algorithmic footings.
C. Toward AI-Specific Legislation
As its third prong, the AI legal development strategy proposed by this work contemplates that the two other legal development prongs will require augmentation to establish a robust AI governance system focused, particularly, on AI bias and discrimination and the thereby-implicated civil rights and consumer protection concerns.[32] Here, new AI-specific laws are needed. For example, laws are needed to ban and enforce against inappropriate uses of AI-effectuated biometric technology[33] and autonomous weapons that kill people outside human control.[34]
The California legislature has been considering AI-specific civil rights legislation, the Automated Decisions Systems Accountability Act.[35] This Assembly Bill No. 13 would adopt a structure synthesized from Title VI of the federal Civil Rights Act of 1964 and the proposed federal Algorithmic Accountability Act of 2019, typing the former’s receipt of government funds[36] to a requirement under the latter’s definition of high risk AI systems to assess the impacts of discrimination and other algorithmic harms from the uses of those systems.[37]
Finally, because AI often operates without the restraint of territorial borders, it will be important to look to harmonize AI governance where possible and appropriate. Just as AI standards as soft law are informed globally, the legal development prong to enact new AI-specific laws likewise should look to other jurisdictions for points of commonality and coordination.[38]
II. Conclusion
By pursuing the multi-pronged legal development strategy proposed in this work, informed, actionable, and enforceable AI governance can be quickly established to appropriately guide and control government and other uses of human-impactful artificial intelligence systems. With this approach, individuals, communities, the rule of law, and democracy will be better protected with greater and enforceable algorithmic justice.
Emile Loza de Siles is Assistant Professor of Law at Duquesne University School of Law. Her scholarship centers on artificial intelligence (AI) and law emphasizing interdisciplinary science and technology topics, social justice, and critical theory.
Emile joined the legal academy in 2019 after sixteen years in intellectual property and technology practice in her firm, Technology Law Group, and in the Office of General Counsel, U.S. Department of Commerce. Her private clients have included HP, Cisco Systems, Accenture, and numerous other technology-driven organizations and innovators.
She clerked for the Honorable Sergio A. Gutiérrez of the Idaho Court of Appeals and the Honorable Sheila F. Anthony of the U.S. Federal Trade Commission.
Emile holds a technology BS, MBA, and her JD from The George Washington University School of Law. She holds a cybersecurity strategy graduate certificate from Georgetown University and has another in data science underway with Harvard University.
[1] See Raj Ramesh, What Is Artificial Intelligence, YouTube (Aug. 13, 2017), https://www.youtube.com/watch?v=2ePf9rue1Ao [https://perma.cc/9KYW-4RGP]; Emile Loza de Siles, AI, on the Law of the Elephant: Toward Understanding AI, 69 Buff. L. Rev. (forthcoming 2021).
[2] See, e.g., Argonne Nat’l Lab’y, Using AI to Enhance Robotic Surgery and Improve Patient Outcomes (last visited Nov. 16, 2021), https://www.anl.gov/cels/using-ai-to-enhance-robotic-surgery-and-improve-patient-outcomes [https://perma.cc/ZC4F-2ZVT].
[3] See, e.g., Nat’l Aeronautics & Space Admin., NASA’s Mars Rover Drivers Need Your Help (June 12, 2020), https://mars.nasa.gov/news/8689/nasas-mars-rover-drivers-need-your-help/ [https://perma.cc/8K27-YEFY].
[4] See, e.g., Dominic Rushe, Tesla’s Autopilot Faces US Investigation After Crashes with Emergency Vehicles, The Guardian (Aug. 16, 2021), https://www.theguardian.com/technology/2021/aug/16/teslas-autopilot-us-investigation-crashes-emergency-vehicles [https://perma.cc/7C7P-BM7Z].
[5] See, e.g., Ben Husch & Anne Teigen, Nat’l Conf. of State Legis’s, Regulating Autonomous Vehicles 25 Legis Brief – (Apr. 2017), https://www.ncsl.org/documents/legisbriefs/2017/lb_2513.pdf [https://perma.cc/S6BG-GA28].
[6] But see, e.g., AI Now Institute, New York University, Litigating Algorithms: Challenging Government Use of Automated Decision Systems (Sept. 2018), https://ainowinstitute.org/litigatingalgorithms.pdf [https://perma.cc/Z5E5-AWBE].
[7] See, e.g., Hanson Robotics, Press Release: Hanson Robotics Announces “Little Sophia,” New Educational STEM Companion for Kids 7 to 13 (Jan. 30, 2019), https://www.hansonrobotics.com/hanson-robotics-announces-little-sophia-new-educational-stem-companion-for-kids-7-to-13/ [https://perma.cc/HQY8-SESX].
[8] See, e.g., Will Douglas Heaven, Predictive Policing Algorithms Are Racist. They Need to Be Dismantled, MIT Tech. Rev. (July 17, 2020), https://www.technologyreview.com/2020/07/17/1005396/predictive-policing-algorithms-racist-dismantled-machine-learning-bias-criminal-justice/ [https://perma.cc/NW5P-KEAN]; Mack DeGeurin, Stop-And-Frisk and Broken Windows Haven’t Gone Away – They’ve Moved Online, Intelligencer (June 12, 2018), https://nymag.com/intelligencer/2018/06/how-predpol-and-nypd-create-digital-stop-and-frisk.html [https://perma.cc/CK2U-JL94].
[9] See, e.g., Sea Hunter: Inside the US Navy’s Autonomous Submarine Tracking Vessel, Naval Tech. (last updated Jan. 30, 2020, 12:23 PM), https://www.naval-technology.com/features/sea-hunter-inside-us-navys-autonomous-submarine-tracking-vessel/ [https://perma.cc/RDE5-GW5W].
[10] See, e.g., Joe Hernandez, A Military Drone with a Mind of Its Own Was Used in Combat, U.N. Says, Nat’l Pub. Radio (June 1, 2021, 3:09 PM), https://www.npr.org/2021/06/01/1002196245/a-u-n-report-suggests-libya-saw-the-first-battlefield-killing-by-an-autonomous-d [https://perma.cc/R9QC-RJWA].
[11] See, e.g., Aaron Boyd, Biometrics in Action, Special Rep’t., Nextgov (Dec. 16, 2020), https://www.nextgov.com/emerging-tech/2020/12/biometrics-action/170771/ [https://perma.cc/9KSQ-QVEL].
[12] See Artificial Intelligence (AI) Market Size, Fortune Bus. Insights (July 2020), https://www.fortunebusinessinsights.com/industry-reports/artificial-intelligence-market-100114 [https://perma.cc/3UYP-CTBJ]; Artificial Intelligence in Government: Global Markets 2020–2025, Business Wire (Aug. 31, 2020, 02:28 PM), https://www.businesswire.com/news/home/20200831005637/en/Artificial-Intelligence-in-Government-Global-Markets-2020-2025—ResearchAndMarkets.com [https://perma.cc/A6J3-BSDQ].
[13] See, e.g., Emile Loza de Siles, The Impossibility of Proof: State Legislation as Critical to Establishing Disparate Treatment by Artificial Intelligence (June 7, 2021) (on file with author); Robert Brauneis & Ellen P. Goodman, Algorithmic Transparency for the Smart City, 20 Yale J.L. & Tech. 103, 114–18 & 126–28 (2018); Deirdre K. Mulligan & Kenneth A. Bamberger, Saving Governance-By-Design, 106 Cal. L. Rev. 697 (2018); Cathy O’Neil, Weapons of Math Destruction: How Big Data Increases Inequality and Threatens Democracy 30–31 (2016); Lawrence Lessig, Codes and Other Laws of Cyberspace 60 (1999) (citation omitted) (“If code is law, then . . . ‘control of law is power’.”).
[14] Tim Cook, CEO, Apple Inc., Keynote Address at the 40th Int’l Conf. of Data Protection & Privacy Commissioners: Debating Ethics: Dignity and Respect in Data Driven Life, YouTube (Oct. 24, 2018), https://www.youtube.com/watch?v=kVhOLkIs20A [https://perma.cc/J625-NQKA].
[15] See, e.g., Thomas Hellström, Virginia Dignum & Suna Bencsch, Bias in Machine Learning: What Is It Good (and Bad) For?, Cornell Univ. ArXiv.org 1, 2 (Apr. 1, 2020) (citation omitted); Patrick Grother, Mei Ngan & Kayee Hanaoka, Face Recognition Vendor Test (FRVT) Part 3: Demographic Effects, Rep’t No. NISTIR 8280, Nat’l Inst. of Standards & Tech. (Dec. 2019), https://www.nist.gov/publications/face-recognition-vendor-test-part-3-demographic-effects [https://perma.cc/B8TL-CTJC]; Sandra G. Mayson, Bias In, Bias Out, 128 Yale L.J. 2218 (2019); James A. Allen, The Color of Algorithms: An Analysis and Proposed Research Agenda for Deterring Algorithmic Redlining, 46 Fordham Urb. L.J. 219 (2019); Stephanie K. Glaberson, Coding over the Cracks: Predictive Analytics and Child Protection, 46 Fordham Urb. L.J. 307 (2019); Richard F. Lowden, Risk Assessment Algorithms: The Answer to an Inequitable Bail System?, 19 N.C. J.L. & Tech. On. 221 (2018); Solon Barocas & Andrew D. Selbst, Big Data’s Disparate Impact, 104 Calif. L. Rev. 671 (2016); Jeffrey L. Vagle, Tightening the OODA Loop: Police Militarization, Race, and Algorithmic Surveillance, 22 Mich. J. Race & L. 101 (Fall 2016); Cathy O’Neil, Weapons of Math Destruction 1–31 & 179–97 (2016) (Introduction, Bomb Parts & Targeted Citizen chapters).
For example, as to the Correctional Offender Management and Profiling Alternative Sanctions (“COMPAS”) system widely used for criminal risk, violence, substance abuse recidivism, and other purported predictions, see Melissa Hamilton, The Biased Algorithm: Evidence of Disparate Impact on Hispanics, 56 Am. Crim. L. Rev. 1553, 1553–54, 1557–58 (2019); Anne L. Washington, How to Argue with an Algorithm: Lessons from the COMPAS-Propublica Debate, 17 Colo. Tech. L.J. 131 (2018); Julia Angwin et al., Machine Bias: There’s Software Used across the Country to Predict Future Criminals. And It’s Biased against Blacks, ProPublica (May 23, 2016), https://www.propublica.org/article/machine-bias-risk-assessments-in-criminal-sentencing [https://perma.cc/4CA4-AAZE]; see also State v. Loomis, 881 N.W.2d 749 (Wis. 2016) (challenge to sentencing use of COMPAS).
[16] See, e.g., Tate Ryan-Mosley, The New Law Suit Shows that Facial Recognition Is Officially a Civil Rights Issue, MIT Tech. Rev. (Apr. 14, 2021), https://www.technologyreview.com/2021/04/14/1022676/robert-williams-facial-recognition-lawsuit-aclu-detroit-police/ [https://perma.cc/9F7Q-UN4Q].
[17] See, e.g., Jonah Engel Bromwich, Daniel Victor & Mike Isaac, Police Use Surveillance Tool to Scan Social Media, A.C.L.U. Says, N.Y. Times (Oct. 11, 2016), https://www.nytimes.com/2016/10/12/technology/aclu-facebook-twitter-instagram-geofeedia.html?ref=oembed [https://perma.cc/MW3W-983A].
[18] See, e.g., Independent High-level Expert Group on Artificial Intelligence, European Comm’n, Ethics Guidelines for Trustworthy AI, (Apr. 8, 2019), https://ec.europa.eu/futurium/en/ai-alliance-consultation/guidelines [https://perma.cc/X2HP-LKC8].
[19] See, e.g., Nat’l Inst. of Science & Tech, Kicking Off NIST AI Risk Framework (Oct. 19–21, 2021), https://www.nist.gov/news-events/events/2021/10/kicking-nist-ai-risk-management-framework [https://perma.cc/EV95-PDZS]; Carlos Ignacio Gutiérrez & Gary Marchant, Center for Law, Science & Innovation, Arizona State University Sandra Day O’Connor School of Law, Soft Law Governance of Artificial Intelligence 3 (2021), https://papers.ssrn.com/sol3/papers.cfm?abstract_id=3855171 [https://perma.cc/LLY9-CUM8]; Jessica Fjeld et al., Berkman Klein Ctr. for Internet & Soc’y at Harvard Univ., Principled Artificial Intelligence: Mapping Consensus in Ethical and Rights-based Approaches to Principles for AI, Res. Pub. No. 2020–1 (Jan. 15, 2020), https://cyber.harvard.edu/publication/2020/principled-ai [https://perma.cc/WX4E-W3S5]; Inst. of Electrical & Electronics Engineers (“IEEE”) Standards Ass’n, Ethically Aligned Design: A Vision for Prioritizing Human Well-Being with Autonomous and Intelligence Systems, ver. 2 (2017) [hereinafter, “Ethically Aligned Design”] https://standards.ieee.org/industry-connections/ec/autonomous-systems.html [https://perma.cc/U7S7-EPFW].
[20] See, e.g., IEEE Standards Ass’n, IEEE 7010-2020 – IEEE Recommended Practice for Assessing the Impact of Autonomous and Intelligent Systems on Human Well-Being, https://standards.ieee.org/standard/7010-2020.html [https://perma.cc/EC2J-K85M]; id., in 1 IEEE Std. 7010–2020 (May 1, 2020), doi: 10.1109/IEEESTD.2020.9084219 & available via IEEE Xplore, https://ieeexplore.ieee.org/document/9084219 [https://perma.cc/53HX-ESYL] and IEEE SA Standards Store, https://www.techstreet.com/ieee/standards/ieee-7010-2020?product_id=2095699 [https://perma.cc/LXZ8-2NNC]; Ethically Aligned Design, supra.
[21] See José E. Alvarez, International Organizations as Law-makers 248–49 (2005) (citation omitted).
[22] See Michael Kearns & Aaron Roth, The Ethical Algorithm: The Science of Socially Aware Algorithm Design 57–93 (2020).
[23] See Gary Marchant, “Soft Law” Governance of Artificial Intelligence, UCLA AI Pulse (Jan. 25, 2019), https://aipulse.org/soft-law-governance-of-artificial-intelligence/ [https://perma.cc/8YJL-A74V]; see, e.g., Nat’l Inst. of Standards & Tech., U.S. Leadership in AI: A Plan for Federal Engagement in Developing Technical Standards and Related Tools (Aug. 9, 2019), https://www.nist.gov/news-events/news/2019/08/plan-outlines-priorities-federal-agency-engagement-ai-standards-development [https://perma.cc/R87P-5MJ9].
[24] See, e.g., Nat’l Inst. of Standards & Tech. (“NIST”), Framework for Improving Critical Infrastructure Cybersecurity, ver. 1.1 (Apr. 16, 2018), [hereinafter, “Cybersecurity Framework”] https://doi.org/10.6028/NIST.CSWP.04162018 [https://perma.cc/7CJH-CTSK]; Andrea Arias, The NIST Cybersecurity Framework and the FTC, Fed. Trade Comm’n (“FTC”) (Aug. 31, 2016, 2:34 PM), https://www.ftc.gov/news-events/blogs/business-blog/2016/08/nist-cybersecurity-framework-ftc [https://perma.cc/P7MG-ZL96]; Emile Loza de Siles, Cybersecurity Law & Emerging Technologies: The Federal Trade Commission, Reasonable Security Measures, and IoT, IEEE Future Directions: Tech. Policy & Ethics (May 2017), https://cmte.ieee.org/futuredirections/tech-policy-ethics/may-2017/cybersecurity-law-and-emerging-technologies-part-1/ [https://perma.cc/4DYK-8D9U].
[25] See supra note 15 [COMPAS, etc.] & accompanying text.
[26] See, e.g., The Algorithmic Accountability Act of 2019, H.R. 2231 & S. 1108 (mirror bills introduced Apr. 10, 2019) (introduced with no further action after referrals to U.S. House & Senate committees).
[27] A recent federal statute was enacted to commence a national focus on artificial intelligence, however. See National Artificial Intelligence Initiative Act of 2020, Div. E William M. (Mac) Thornberry National Defense Authorization Act for Fiscal Year 2021, Pub. L. No. 116–283 (Jan. 1, 2021), https://www.congress.gov/bill/116th-congress/house-bill/6395?__cf_chl_jschl_tk__=pmd_QgChBrhCJjAUZGNi9rc5.52UYORPq0ZeKjEi4M0kTTA-1635801474-0-gqNtZGzNAnujcnBszQdl [https://perma.cc/AKN5-H4Z3].
[28] See, e.g., U.S. Housing & Urban Dev. Dep’t, Rule: Implementation of the Fair Housing Act’s Disparate Impact Standard, 85 Fed. Reg. 60,288 (2020). As to predictive AI models, “HUD expects that there will be further development in the law in the emerging technology areas of algorithms, artificial intelligence, machine learning and similar concepts. Thus, it is premature at this time to more directly address algorithms.” Id.
[29] See Andrew Smith, Director, Bureau of Consumer Protection, Fed. Trade Comm’n, Using Artificial Intelligence and Algorithms (Apr. 8, 2020), https://www.ftc.gov/news-events/blogs/business-blog/2020/04/using-artificial-intelligence-algorithms [https://perma.cc/3KKD-S4XK] (“We believe that our experience, as well as existing laws, can offer important lessons about how companies can manage the consumer protection risks of AI and algorithms.”).
[30] See, e.g., U.S. Food & Drug Admin., Artificial Intelligence and Machine Learning in Software as a Medical Device Action Plan (Sept. 22, 2021), https://www.fda.gov/medical-devices/software-medical-device-samd/artificial-intelligence-and-machine-learning-software-medical-device [https://perma.cc/M497-AEW6].
[31] See, e.g., Fed. Trade Comm’n, Protecting Consumer Privacy in an Era of Rapid Change: Recommendations for Businesses and Policymakers 22–34 (Mar. 26, 2012), https://www.ftc.gov/reports/protecting-consumer-privacy-era-rapid-change-recommendations-businesses-policymakers [https://perma.cc/Q38G-QUVQ].
[32] See, e.g., Algorithmic Accountability Act of 2019, S. 1108 & H.R. 2231, 116th Cong. (identical bills introduced Apr. 10, 2019).
[33] See, e.g., City and County of San Francisco, Administrative Code – Acquisition of Surveillance Technology, Ordinance No. 107-19, File No. 190568 (as amended May 29, 2019); Thomas Burri & Fredrik von Bothmer, The New EU [Proposed] Legislation on Artificial Intelligence: A Primer 2-3 (Apr. 21, 2021), https://papers.ssrn.com/sol3/papers.cfm?abstract_id=3831424 [https://perma.cc/F83R-JNGX].
[34] Hernandez, supra note 10.
[35] Automated Decisions Systems Accountability Act, Assemb. 13, 2021–2022 Leg., Reg. Sess. (Cal. 2021).
[36] 42 U.S.C. § 2000d.
[37] See Algorithmic Accountability Act of 2019, S. 1108 & H.R. 2231, supra, § 2(7).
[38] See, e.g., Singapore Personal Data Protection Comm’n, Model Artificial Intelligence Governance Framework (2d ed., Jan. 2020), https://www.pdpc.gov.sg/Help-and-Resources/2020/01/Model-AI-Governance-Framework [https://perma.cc/TPE6-XN9E]; Independent High-level Expert Group on Artificial Intelligence, European Comm’n, Ethics Guidelines for Trustworthy AI, (Apr. 8, 2019), https://ec.europa.eu/futurium/en/ai-alliance-consultation/guidelines [https://perma.cc/AQH8-SEWV].
