Require AI companies to disclose training data to the federal government on request
Artificial intelligence
Owners remove an obstacle nothing else removes. Others must also act.
Why shareholder democracy is necessary
Only Congress can compel disclosure to a federal agency. The companies hold the training data and can disclose it voluntarily, so ownership pressure moves the underlying outcome. Shareholder proposals seeking AI transparency reports were filed at five major media and technology companies by the AFL-CIO Equity Index Funds in 2023, and ICCR's digital rights program presses technology companies for accountability in how AI systems are built and deployed. The companies' resistance to disclosure mandates is now the documented obstacle. An industry coalition including the Software and Information Industry Association formally opposed California's AB 412 training data disclosure bill as infeasible, and xAI sued in December 2025 to strike down California's Training Data Transparency Act, arguing that revealing its training data would destroy trade secrets. Majorities of both parties already support disclosure, and nothing stands between that support and the outcome except the industry's own political and legal resistance. We judge that influence to be the linchpin: owners who redirect what their companies lobby and litigate for can release the block.
How this was scoredNecessary
Necessary. Owners remove an obstacle nothing else removes. Others must also act.
We have not yet written down the two answers behind this rating. Recording them on every item, so a reader can check the reasoning and not only the conclusion, is work in progress.
Every rating is the higher of two answers, one for what companies do and one for what they spend to stop the outcome, because the same vote reaches both. See the full method.
Sources
- Common Ground of the American People: Artificial Intelligence, Program for Public Consultation / Voice of the People, 2026 (study February 2024)
- Artificial Intelligence Transparency will Fight the Dehumanization of the American Workforce, says AFL-CIO Equity Index Funds Shareholder Proposal, AFL-CIO, 2023-10-19
- Digital Rights and AI Accountability, Interfaith Center on Corporate Responsibility, accessed 2026-08-28
- SIIA Joins Coalition Letter Opposing California AB 412: AI Training Data Disclosure Requirements, Software & Information Industry Association, 2026-06-22
- xAI Challenges California's Training Data Transparency Act, Goodwin, 2026-01
What civil society organizations are helping owners on this
Filed proposals at Microsoft (2024) and Apple (2025) asking for reports on the ethical sourcing and use of external data in training AI systems.
Engages technology companies to adopt accountability and human rights due diligence for the development and use of artificial intelligence.
Digital Rights and AI Accountability, Interfaith Center on Corporate Responsibility
6 campaigns on record
Publish an annual report assessing the risks of unethical data acquisition and use in AI development, including through partners such as OpenAI, with mitigation steps and effectiveness measures.
Result: 12 percent support at the February 25, 2025 meeting per Sustainalytics
Report on the risks of unethical or improper use of external data in the development and training of the company's AI offerings and the safeguards in place.
Result: 36 percent support at the December 10, 2024 meeting per Sustainalytics
Publish a report on how the company uses artificial intelligence, what ethical guidelines it follows to protect workers, customers and the public, and how the board oversees it.
Result: Not approved at the June 6, 2024 annual meeting, with 144,328,782 votes for and 189,038,209 against. Governance Intelligence reports the proposal drew 43.3 percent of the votes cast and was filed through Segal Marco Advisors.
we found no documented change
Publish a report on how the company uses artificial intelligence, what ethical guidelines it follows to protect workers, customers and the public, and how the board oversees it.
Result: Withdrawn after the company agreed to provide more transparency about its use of artificial intelligence.
Comcast agreed to provide additional disclosure about its use of artificial intelligence.
Publish a report on how the company uses artificial intelligence, what ethical guidelines it follows to protect workers, customers and the public, and how the board oversees it.
Result: Withdrawn before the April 3, 2024 meeting after the company committed to greater disclosure about its use of artificial intelligence.
Disney agreed to expand what it discloses about its use of artificial intelligence.
Publish a report on how the company uses artificial intelligence, what ethical guidelines it follows to protect workers, customers and the public, and how the board oversees it.
Result: Voted down at the February 28, 2024 annual meeting with 37.5 percent support. The SEC had refused Apple permission to leave the proposal off the ballot, saying it raised a question that goes beyond ordinary business.
we found no documented change
Related subjects on other maps
How this rating was made
Every item on every one of our maps is read against one question: how far does this depend on the people who own companies voting their values? We answer it twice, once for what companies do and once for what they spend to stop the outcome, and take whichever answer is higher, since the same vote reaches both. The score carries its reasoning and its sources so that a reader can check it rather than take it.
- Pivotal. Owners voting their values can deliver most of the outcome.
- Necessary. Owners remove an obstacle nothing else removes, and others must also act.
- Helpful. Owners voting their values help, and others carry most of it.
- Independent. This moves without owners. Other levers carry it.
The advocacy record on this page holds only what a source we opened says happened. Where we searched and found nothing, the page says so rather than leaving a silence. Where a campaign names a filer the source does not name, it says that too. Ratings are ours; the list of subjects is Common Ground of the American People's, and we link to their original.
How we score dependency, in full.