Artificial Intelligence Disclosure

Jul 11, 2024
Sustainability Reporting
Labrador

Introduction

It is a rare company – if any – that has not considered the implications of artificial intelligence¹ on its industry, its operations and its business model. The ramifications of fast-developing technology for a business’s sustainability are manifold, and companies need to consider whether and how those ramifications warrant discussion in their sustainability reports. This thought piece explores the intersection of AI and ESG and issues and topics that companies might consider in developing their next sustainability report. It is meant to provoke thought rather than provide answers, and as such the requirements of international disclosure regulations, standards and frameworks are not addressed here (and should be considered).

Disclosure Principles

In the US, in the absence of regulatory mandates, every company’s sustainability report is bespoke. Disclosures in sustainability reports tend to be driven by the level of integration of sustainability into a company’s strategy, the depth of a company’s measurable commitment to sustainability, and the intended audience. However, the overwhelming use of voluntary international standards and frameworks provide some baseline of disclosure approach. Labrador’s benchmarking of sustainability reports published in 2023 by a cross-industry mix of 100 companies from the Standard & Poor 500 found that:

76% reported with reference to or in accordance with GRI, which focuses on the material² impact a company has on the economy, the environment, and people and society;
82% reported following the TCFD standards, which focuses on the material financial impact to the company of climate matters; and
90% reported on the SASB standards for their industries, again focusing on material financial impact of environmental and social matters.

Assuming that the practices of the benchmarked companies are generally adopted by companies outside that group, it is reasonable to infer that most sustainability reports contain disclosures of matters that are significant to the company based on their financial impact or based on the company’s impact of the world around it. In the context of AI then, a starting point for considering what to disclose in the report is three-pronged: what do your stakeholders want to know about/what is important to them regarding the company’s interaction with AI; what impact does AI have on your operations, strategy and results; and what impact does your use of AI have on people, the environment and the economy? Companies reporting in alignment with GRI will recognize this as similar to the considerations for determining material topics set out in GRI 3.

At the same time, reporting companies need to think about disclosure rules and regulations as they discuss AI capabilities and the role in their business. SEC Chair Gary Gensler has cautioned against “AI washing” and emphasized that companies must be “truthful about its use of AI and associated risk.”³

This article makes no attempt to distinguish the various types of artificial intelligence.
Unless otherwise stated, references in this thought piece to “material” refer to the definition given to that term in the relevant regulation, standard or framework, and not to the definition of materiality under US federal securities law.
https://www.sec.gov/news/speech/gensler-ai-021324

Below are some issues and matters to evaluate through this lens as companies consider the content of their next sustainability report.

AI and Governance

Governance, loosely defined, is how the company makes and manages decisions, maintains integrity and compliance, and provides transparency to its stakeholders. The use of a relatively new and rapidly-developing technology like AI involves governance considerations on many levels. According to a Deloitte survey of over 2,800 executives across six industries and 16 countries in late 2023, while the vast majority of respondents (79%) expected generative AI to drive substantial transformation within their organization and industry over the next three years, only 25% said they were highly prepared to address governance and risk concerns (with 41% reporting that their companies were only slightly or not at all prepared). The speed of AI evolution and adoption (or pressure to adopt) may outpace more methodical governance decisions.

Questions regarding governance that companies may want to consider and address in sustainability reports include:

What governance structures are in place around how AI is evaluated, used and monitored in the company’s operations, products and services?
1. How are AI risks integrated into the company’s risk management system?⁴
2. How are AI risks being incorporated into the crisis management program (including resilience and redundancy)?
3. Does the company have AI Principles governing the ethical use of AI?⁵
How is the impact of AI being assessed, and what kinds of impact — financial and operational impact on the company, or the impact on people, the environment and the economy, or both?
How is the integrity of AI being managed (ethical use and algorithmic bias, intentional or historic/unintentional)⁶
How does the board oversee AI use and its implications? What is the frequency and quality of information from management?
Should AI be a separate board skill? Should there be board proficiency, fluency or expertise?
How are employees educated on how AI works and is used in the business; its benefits and pitfalls; ethical considerations?
How is AI used in creating public disclosures/content?
How is AI helping achieve ESG or transition goals?

The US National Institute of Standards and Technology (NIST) issued an Artificial Intelligence Risk Management Framework in January 2023.
See for example Google, ai.google/responsibility/principles.
See Former SEC Commissioner Kara Stein’s discussion of programmable (intentional), predictable (reckless) and unpredictable harm (AI Governance in the Financial Industry, Stanford Journal of Law, Business and Finance Vol. 27 No. 1, 2022)

AI and Human Capital Management

A key component of the Biden administration’s October 2023 Executive Order on Safe, Secure and Trustworthy Artificial Intelligence was to address job displacement and workforce training. Deloitte’s survey found that only 22% of respondents felt that their companies were highly prepared to address talent concerns, and Deloitte believes that most organizations have yet to substantially address workforce challenges arising from large-scale AI adoption.

Consider what human capital matters are sufficiently significant and ripe to address in the sustainability report (and Form 10-K), perhaps including:

What positive and negative impacts of AI are occurring/anticipated to occur on the workforce? How are those impacts being managed?
Is AI being used to improve the quality of worker life? Is training/upskilling provided?
How is the company engaging/communicating with the workforce to make them feel comfortable about how AI will impact the workplace and address concerns?
How is the company planning for the workforce of tomorrow as a result of AI use?

AI and Strategy/Operations

While the Form 10-K and annual report are the venues for reporting on the details of company operations and strategy, the sustainability report should touch on those topics to put sustainability in context. Consider whether and how the sustainability report should discuss the role of AI in the company’s strategy and operations, so that stakeholders can better understand what sustainability means for the company’s future. Matters to consider include:

How is AI changing how the business will be operated?
How is AI changing the products and services offered by the company?
What are the financial implications – capital needs, costs, cost savings, revenue, etc. – of AI?

AI and Data Privacy and Security

The Biden executive order also focused on protecting the privacy of personal data. AI may similarly pose a threat to business data. In connection with the company’s cybersecurity efforts, consider whether it is appropriate to discuss plans, programs and actions designed to protect employee, customer, and proprietary data when interacting with AI.

AI and the Environment

The use of AI presents environmental issues ranging from the natural resource impact of chip production, to emissions and water use resulting from the computing power needed to train and run AI, to the environmental impact of the products brought to market as a result of AI. US legislators are concerned about this issue; for example, in February 2024, the House and the Senate introduced identical bills – the Artificial Intelligence Environmental Impacts Act of 2024 – seeking to develop standards (through NIST) to measure and report the full range of AI’s environmental impact.

Consider how the impact of AI on environmental sustainability should be discussed in the sustainability report.

Is the company able to calculate Scope 1 and Scope 3 emissions (upstream and downstream) resulting from its use of AI?
Can/will the company assess the other environmental impacts arising from the use of AI?
How will adverse impacts be mitigated? Does the company expect environmental benefits from its use of AI?

Shareholder Proposals Concerning AI

As quickly as ChatGPT stormed in to make AI accessible to the masses, so did shareholder proposals seeking more disclosure about how companies are using, and evaluating the impact of, AI. In the last two years, those proposals mainly focused on four concerns: (1) the risk of misinformation generated via AI; (2) transparency about the ethical parameters around AI use; (3) the impact on human rights of AI-driven targeted advertising; and (4) transparency about board oversight of AI. The most prominent of the AI-related shareholder proposals are set out below (results rounded to nearest 1%).

While none of these proposals passed, the flurry of AI proposals indicates that investors want assurances regarding the use and impact of AI. The proposals requesting reports about ethical company use of AI garnered the most support, with 43% and 38% of shares voting to approve such a report at Netflix’s and Apple’s 2024 annual meetings, respectively.

Company (Proponent)	Date	Subject of Proposal	Disposition
Misinformation risk report - Prepare and publicly disclose a “misinformation risk report” assessing risks to the company and public welfare presented by the company’s role in facilitating misinformation and disinformation disseminated or generated via artificial intelligence, and steps the company plans to remediate those harms
Meta (Arjuna/OpenMIC)	Early 2024		AGM 5/29/24; 17% approval (note: 61% voting shares owned by CEO)
Alphabet (Arjuna/OpenMIC)	Early 2024		AGM 6/7/24; 18% approval (note: 53% voting shares owned by directors and executive officers)
Microsoft (Arjuna/OpenMIC)⁷	Late 2023		AGM 12/7/23; 21% approval
Transparency report - Prepare and publicly disclose a “transparency report” regarding its use of AI, including ethical guidelines
Netflix (AFL-CIO)	Mid 2024		AGM 6/6/24; 43% approval
Warner Brothers Discovery (AFL-CIO and various NYC retirement systems)	Mid 2024		AGM 6/3/24; 24% approval
Apple (AFL-CIO)	Early 2024		SEC did not concur with no-action request AGM 2/24/24; 38% approval
Paramount Global (two NYC retirement systems)	Early 2024		SEC did not concur with no-action request AGM 6/4/24; 2% approval (note: 77% of voting shares controlled by Redstone affiliate)
Disney (AFL-CIO)	Early 2024	Also requested disclosure on board’s role in overseeing AI use	SEC did not concur with no-action request Proposal withdrawn, in response to commitment to enhance disclosures about ethical use of AI
Comcast (AFL-CIO)	Early 2024		Proposal withdrawn, in response to commitment to enhance disclosures about ethical use of AI
United Health Group (SHARE)	Early 2024		Withdrawn. UNH agreed to produce such a report in mid-2024. See enhanced disclosure in 2023 Sustainability Report.
Human impact of AI report
Republic Services (Teamsters)	Early 2024	Prepare and publicly disclose a report regarding the impact of its climate change strategy on relevant stakeholders, consistent with certain “Just Transition” guidelines that include impact of AI on workforce.	SEC did not concur with no-action request AGM 5/23/24; 15% approval
Meta (Mercy Investment Service)	Mid 2023 Early 2024	Prepare and publicly disclose report on human rights impact of AI-driven targeted advertising	AGM 5/31/23; 17% approval (note: 61% voting shares owned by CEO) AGM 5/29/24; 14% approval (note: 61% voting shares owned by CEO)
Chipotle (Teamsters)	Early 2024	Prepare and publicly disclose report addressing social implications on workforce of AI and other advanced technologies	AGM 6/6/24; 18% approval
Alphabet (Shareholder Association for Research & Education (SHARE))	Early 2024	Prepare and publicly disclose report on human rights impact of AI-driven targeted advertising	AGM 6/7/24; 19% approval (note: 53% voting shares owned by directors and executive officers)
Board oversight – amend governance mandates to explicitly include AI
Amazon (AFL-CIO)	Mid 2024	Establish board committee focused on human rights risks of AI	AGM 5/22/24; 10% approval (note: 11% voting shares owned by directors and executive officers)
Alphabet (Trillium)	Early 2024	Amend board charters to explicitly cover oversight of AI and AI Principles	AGM 6/7/24; 7% approval (note: 53% voting shares owned by directors and executive officers)
Toronto Dominion Bank (MEDAC)	Early 2023	Amend governance mandates to include AI ethics, and develop AI ethical code covering, inter alia, regulating algorithms to overcome bias	Withdrawn. TD agreed to delineate specific AI oversight in Risk Committee charter
Royal Bank Canada (MEDAC)	Early 2023	Amend governance mandates to include AI ethics, and develop AI ethical code covering, inter alia, regulating algorithms to overcome bias	Withdrawn. RBC will provide enhanced disclosure of board AI oversight in next annual report
Other
Alphabet (Trillium)	Mid 2023	Increased disclosure on algorithmic systems	AGM 6/2/23; 17% approval (note: 53% voting shares owned by directors and executive officers)

Trivia note: this was filed by Ajuna Capital on behalf of Krist Novoselic, former bassist for Nirvana

Conclusion

Determining what to discuss about AI in a sustainability report is not fundamentally different than the analysis for any other topic that relates to the future health of the company, its workforce, and the communities, environment and economies that it touches. What is different, however, is the lightning speed with which AI has become almost an imperative for companies. As a result, companies may want to, with due haste, thoughtfully assess the impact of AI on all corners of the business and outwardly on the physical and social spaces in which the company operates. Since this is an evolving area, staying abreast of disclosure practices will be important.