The Commission’s economic analysis does not sufficiently account for the information shareholder proposals generate on emerging risks that are not yet addressed by disclosure rules or industry regulation. Artificial intelligence is a topical example of that. Amazon, Alphabet, Meta, Microsoft and SpaceX alone expect to spend between $803 billion and $828 billion on capital expenditures in 2026, driven largely by AI infrastructure,[1] yet of 3,048 companies in the Russell 3000 and S&P 500, only 8% disclose board-level oversight of AI.[2] Investors have responded to that gap through Rule 14a-8 , bringing 66 AI-related shareholder proposals to a vote at U.S. companies since the 2023 proxy season.[3] Without a standardized disclosure framework or established industry best practice, these proposals have been one of the few ways investors could ask boards to provide additional information and be held accountable for AI-related risks. Rescinding Rule 14a-8 would close that channel just as AI risks become more concentrated in investment portfolios.
ISS STOXX characterizes this as a structural gap, with companies deploying AI faster than they are building the oversight needed to manage its risks.[4] Its review found that only 9% of the same companies reported having an established AI policy.[5] Given the pace of investment, the national security stakes of a strong domestic AI supply chain, and the outsized effect AI risks could have on investor portfolios, this absence of disclosure leaves shareholders with little basis for judging how boards are managing the technology.
Tech founders have themselves acknowledged that governance and regulation around AI have not caught up with the technology. Dario Amodei, the CEO of Anthropic, explains that while Anthropic tests its models and publishes its evaluations, these practices are “broadly voluntary,” and no federal law requires AI companies to disclose their models’ capabilities or to take meaningful steps to reduce risk. As he put it, “Some companies can simply choose not to.”[6] Independent assessments point the same way. Stanford’s 2026 AI Index found that responsible AI practices are not keeping pace with AI capability, with documented AI incidents rising from 233 in 2024 to 362 in 2025,[7] and a United Nations scientific panel concluded in July 2026 that regulation is lagging the technology.[8] Boards and management teams will likely need time to build the expertise to oversee these risks as well. In the meantime, a low-cost way of raising emerging risks, and a channel for investors to request additional information, should continue to exist.
While federal and global regulation is still catching up, many aspects of AI-related operations and activities across the AI supply chain remain exposed to litigation, reputational and financial risks. These risks are no longer hypothetical, and market reactions to developments in this space can be highly volatile. On November 5, 2026, two days after the U.S. midterm elections and amid strong community pushback against data centers, fears of new restrictions triggered a broad selloff of AI-linked stocks. The S&P 500 fell 4.4% and the Nasdaq Composite fell 5.5%, with the Nasdaq 100 alone losing about $2.5 trillion in market value, as Goldman Sachs estimated that states whose newly elected governors had pledged to stop or pause data center construction account for about 45% of planned U.S. data center capacity.[9] This demonstrates that investors, while optimistic about the opportunities for growth that AI brings, continue to monitor these risks closely and disclosure on how management is approaching these risks is a material issue for investors.
These risks are only expected to grow. With Anthropic having filed for an initial public offering in October 2026 that could raise about $75 billion, and OpenAI preparing its own listing,[10] more AI companies will list, more companies will integrate AI into their operations, and AI risks will become further concentrated in the portfolios of everyday shareholders, including pension funds and retirement accounts. In that context, good governance will be critical. Additionally, given the outsized investment in AI and its relevance to the economy right now, the fact that more than 90% of companies do not disclose board-level oversight of AI shows that investors do not have sufficient information on how companies are approaching this rapidly evolving technology.
The growth in investor concern around AI risks has been reflected on corporate proxy ballots over the last few years with 66 AI-related proposals being voted on from the 2023 proxy season, the first after ChatGPT’s public release.[11] And while the number of environmental and social proposals on ballots fell sharply in 2025, ISS found that AI-related proposals did not decline.[12] AI risks that concern a company’s long-term value and financial performance have taken multiple forms:
Environmental. Investors have raised concerns about the reliance on, and development of, large data centers and fabrication facilities, and the water and energy these projects consume. Shareholders asked Amazon in 2025 to report on how its data center expansion affects its climate commitments, and asked Alphabet for enhanced disclosure on its climate goals after the company identified data centers running energy-intensive AI workloads as the primary cause of rising emissions.[13] At Digital Realty, a 2026 proposal from NorthStar Asset Management asked the company to disclose region-level water risk metrics across its data center portfolio, so that the water-intensive nature of its developments does not lead to community pushback or regulatory hurdles that could result in financial losses for shareholders. The proposal received 22.5% support.[14]
Social. The integration of AI tools into company operations will have significant impacts on the workforce, and investors have asked how management plans to ensure that this transition takes place in a just manner, through proposals on working conditions and automation.[15] Investors have also raised broader human rights concerns, including the spread of AI-generated misinformation and disinformation, the human rights impacts of AI-driven targeted advertising, and the effects of AI-enabled content and algorithms on child safety.[16] Data sourcing accountability has also emerged as a concern, with a proposal filed by the National Legal and Policy Center at Microsoft receiving 36.2% support.[17]
Governance. Investors have requested stronger board oversight of AI, including proposals asking Amazon to establish a board committee on AI and Alphabet to add AI oversight to its audit committee charter.[18] Others have asked companies to disclose how they plan to integrate AI into their operations and the risk controls they are putting in place. A set of these proposals, filed by the AFL-CIO at Netflix, Apple, Warner Bros. Discovery and Paramount, asked each company to report on its use of AI, with the proposal at Netflix receiving 43.1% support and the proposal at Apple 37.5%.[19]
As the ISS report notes, many AI-related proposals focus on topics shareholders have historically been concerned about, such as water and emissions-related risks, or the social implications of business operations such as child safety and workforce impacts.[20] But given how ubiquitous AI applications are becoming and how quickly companies are adopting them, these familiar risks stand to be amplified. ISS STOXX reads the continued filing of AI-related proposals as a sign that investors still see material risks as unresolved.
Shareholder proposals have a history of ensuring corporate accountability while regulation catches up. Before Congress required public companies to hold advisory votes on executive pay in 2010, investors had already brought the question to company ballots, with say-on-pay proposals filed at 60 or more companies in 2007, up from seven the year before. Similarly, investors filed proposals asking companies to strengthen policies that recover executive pay from executives to hold them accountable for. misconduct. Several companies adopted better recoupment policies and improved disclosure on the topic well before the SEC adopted its clawback rules in October 2022.[21]
The cost of preserving this channel is modest by the Commission’s own estimate. The Proposing Release estimates that rescission would save companies about $49,000 per proposal, or about $39.6 million a year across all proposals.[22] At that rate, the 66 AI-related proposals that reached a vote from 2023 through 2025 cost companies about $3.2 million in total. That is roughly 0.0004% of what five companies alone expect to spend on capital expenditures in 2026, and a vanishingly small figure next to the $2.5 trillion in market value the Nasdaq 100 lost in a single day when AI policy risk was repriced. The Release itself acknowledges that rescission could reduce the informational spillovers shareholder votes generate, a cost to the broader market and to companies, yet it makes no attempt to measure that cost.[23]
Given AI’s strategic importance, and how quickly its risks can translate into financial losses, keeping a channel of communication open between shareholders and companies is essential. The shareholder proposal remains the most cost-efficient tool for that engagement, especially for smaller shareholders. For an emerging risk of this scale, where disclosure rules and industry practice have not yet caught up, the cost the Release leaves unmeasured is precisely the one that matters most to investors.
[1] Lloyd Lee, “AI hyperscalers are expected to finish the year spending a whopping $828 billion on capex,” Business Insider, republished by Hargreaves Lansdown, October 30, 2026, https://www.hl.co.uk/news/ai-hyperscalers-are-expected-to-finish-the-year-spending-a-whopping-828-billion-on-capex. Figures reflect total capital expenditures; executives identify AI infrastructure as the main driver.
[2] Hinza Zeru and Andrew Schultz, “Mind the Governance Gap: The State of Board Oversight and AI Policy in U.S. Companies,” ISS STOXX, March 3, 2026, https://www.iss-stoxx.com/insights/articles/mind-the-governance-gap-the-state-of-board-oversight-and-ai-policy-in-us-companies/. Findings are based on a review of company disclosures using ISS Governance QualityScore data.
[3] Author’s calculation from Joseph Hong, “A Look at AI-Related Shareholder Proposals at U.S. Companies, 2022-2025,” ISS Governance, December 4, 2025, pp. 4-7. Counts include only proposals that went to a vote; withdrawn or excluded proposals are not captured. ISS defines “AI-related” to include proposals that explicitly address AI as well as proposals at companies where AI is a substantial or core business strategy, and excludes anti-ESG proposals. Years follow ISS’s labels, and the average is a simple, unweighted mean of disclosed support. ChatGPT was publicly released on November 30, 2022.
[4] Zeru and Schultz, supra note 34.
[5] Zeru and Schultz, supra note 34.
[6] Dario Amodei, “Anthropic C.E.O.: Don’t Let A.I. Companies off the Hook,” The New York Times, June 5, 2025, https://www.nytimes.com/2025/06/05/opinion/anthropic-ceo-regulate-transparency.html.
[7] Stanford Institute for Human-Centered Artificial Intelligence, “The 2026 AI Index Report,” April 2026, Top Takeaway 5 and Chapter 3, https://hai.stanford.edu/ai-index/2026-ai-index-report.
[8] “The UN says AI is moving faster than the rules, and it has a report to prove it,” The Next Web, July 2026 (describing the preliminary report of the UN Independent International Scientific Panel on AI, released July 1, 2026), https://thenextweb.com/news/un-warns-ai-outpacing-rules-geneva-dialogue.
[9] David Hollerith, “How midterm elections sent Wall Street from celebration to mayhem in just 24 hours,” Yahoo Finance, November 7, 2026, https://finance.yahoo.com/markets/article/stocks-soared-on-the-democratic-sweep-less-than-24-hours-later-the-ai-trade-cracked-183312401.html (citing Goldman Sachs for the data center capacity estimate). For the Nasdaq 100 market value figure, see Reuters, “Blue Tsunami Wipes out $2.5 Trillion From Nasdaq, as Investors Fret About AI Trade,” republished by GV Wire, November 5, 2026, https://gvwire.com/2026/11/05/blue-tsunami-wipes-out-2-5-trillion-from-nasdaq-as-investors-fret-about-ai-trade/ (Nasdaq 100 fell 6.4%, erasing about $2.5 trillion in market value).
[10] “The IPO Buzz: Anthropic (ANT Proposed) Files its S-1,” IPOScoop, October 26, 2026 (reporting that IPO professionals expect the offering could raise about $75 billion), https://www.iposcoop.com/the-ipo-buzz-anthropic-ant-proposed-files-its-s-1/; “OpenAI aiming for speedy IPO, source says, as market awaits SpaceX filing,” Reuters via BNN Bloomberg, May 20, 2026, https://www.bnnbloomberg.ca/business/2026/05/20/openai-aiming-for-speedy-ipo-source-says-as-market-awaits-spacex-filing/.
[11] Hong, supra note 35, at 4-7.
[12] Hong, supra note 35, at 3.
[13] Amazon.com, Inc., DEF 14A, 2025, Item 7 (impact of data centers on climate commitments), https://www.sec.gov/Archives/edgar/data/1018724/000110465925033442/tm252295-1_def14a.htm; Alphabet Inc., DEF 14A, 2025, Item 7 (enhanced disclosure on climate goals; supporting statement notes the company identifies data centers hosting energy-intensive AI functions as the primary cause), https://www.sec.gov/Archives/edgar/data/1652044/000130817925000511/goog012701-def14a.htm.
[14] Digital Realty Trust, Inc., DEF 14A, 2026, https://www.sec.gov/Archives/edgar/data/1297996/000130817926000296/dlr015307-def14a.htm; NorthStar Asset Management, “Nearly One in Four Digital Realty Shareholders Back NorthStar Proposal on Water Risk,” June 9, 2026, https://northstarasset.com/nearly-one-in-four-digital-realty-shareholders-back-northstar-proposal-on-water-risk/.
[15] See, e.g., Amazon.com, Inc., DEF 14A, 2025, Item 10 (warehouse working conditions), https://www.sec.gov/Archives/edgar/data/1018724/000110465925033442/tm252295-1_def14a.htm; Chipotle Mexican Grill, Inc., DEF 14A, 2024, Item 8 (adoption of automation), https://www.sec.gov/Archives/edgar/data/1058090/000114036124021181/ny20015905x2_def14a.htm.
[16] See, e.g., Microsoft Corporation, DEF 14A for the December 10, 2024 annual meeting, Item 8 (AI-generated misinformation), https://www.sec.gov/Archives/edgar/data/789019/000119312524242883/d858775ddef14a.htm; Alphabet Inc., DEF 14A, 2025, Item 12 (AI-driven targeted advertising), https://www.sec.gov/Archives/edgar/data/1652044/000130817925000511/goog012701-def14a.htm; Meta Platforms, Inc., DEF 14A, 2025, Item 9 (child safety), https://www.sec.gov/Archives/edgar/data/1326801/000132680125000040/meta-20250417.htm.
[17] Microsoft Corporation, DEF 14A for the December 10, 2024 annual meeting, Item 9, https://www.sec.gov/Archives/edgar/data/789019/000119312524242883/d858775ddef14a.htm. Support level from Hong, supra note 35.
[18] Amazon.com, Inc., DEF 14A, 2024, Item 16, https://www.sec.gov/Archives/edgar/data/1018724/000110465924045910/tm2329302d4_def14a.htm; Alphabet Inc., DEF 14A, 2024, Item 11, https://www.sec.gov/Archives/edgar/data/1652044/000130817924000612/lgoog2024_def14a.htm.
[19] Netflix, Inc., DEF 14A, 2024, Item 4, https://www.sec.gov/Archives/edgar/data/1065280/000119312524100913/d521836ddef14a.htm; Apple Inc., DEF 14A, 2024, Item 7, https://www.sec.gov/Archives/edgar/data/320193/000130817924000010/laapl2024_def14a.htm; Warner Bros. Discovery, Inc., DEF 14A, 2024, Item 5, https://www.sec.gov/Archives/edgar/data/1437107/000143710724000102/wbd-20240419.htm; Paramount Global, DEF 14A, 2024, Item 6, https://www.sec.gov/Archives/edgar/data/813828/000119312524105026/d558817ddef14a.htm. Support levels from Hong, supra note 35.
[20] Hong, supra note 35, at 3.
[21] Cooley LLP, “ISS/RiskMetrics’ 2008 Postseason Report Summary,” 2008 (several companies adopted recoupment policies after high 2007 votes, and “a majority of S&P 100 firms now have some type of clawback policy”), https://www.cooley.com/news/insight/2008/iss-riskmetrics-2008-postseason-report-summary; Sidley Austin LLP, “SEC Adopts Final Rules That Will Require the Recovery of Erroneously Awarded Compensation,” November 11, 2022 (rules adopted October 26, 2022), https://sidley.com/en/insights/newsupdates/2022/11/sec-adopts-final-rules-that-will-require-the-recovery-of-erroneously-awarded-compensation.
[22] Release, supra note 4, at 59,938–39 (estimating 107 burden hours per proposal, monetized at $462 per hour, for cost savings of about $49,000 per proposal) and 59,952 (estimating aggregate annual monetized cost savings of about $39.6 million). The comparison in the text is illustrative. It applies the Release’s per-proposal estimate only to the 66 AI-related proposals that went to a vote and does not capture proposals that were withdrawn or excluded.
[23] Release, supra note 4, at 59,944–45.