Demand Side II: Firms, Platforms, and Lobbying

Session 4 · Mon, Oct 5

POLISCI 158 · AI & Politics · Stanford · Fall 2026

Roadmap

In this session
Questions we’ll tackle
Before class
After class

In this session

We examine how control of AI inputs can become political power. We distinguish six channels of firm influence, then compare model access and governance coalitions. Finally, we ask how a model should respond when a lab, a user and a government make competing demands.

Questions we’ll tackle

Why can concentrated resources help firms organize politically?
Which channel connects a firm’s resources to a policy or electoral decision?
What evidence distinguishes participation from successful influence?
What is the difference between open weights, open source AI and hosted access?
Whose interests should a model serve when demands conflict?

Before class

Read

Priority: Amazon’s Antitrust Paradox, Lina M. Khan (2017), Yale Law Journal. Read the introduction and conclusion. Why does Khan argue that a focus on short-term consumer prices misses platform power?
Anthropic’s Economic Policy Framework, Anthropic (June 2026). Read this as a company’s policy proposal. Which problems and solutions does it emphasize?
Stopping Big Tech from Becoming Big AI, Max von Thun and Daniel A. Hanley (2024), Open Markets Institute and Mozilla. Read the introduction and summary on computing power, data, capital and distribution.
Generative Influence, Mike Tanglis and Eileen O’Grady, Public Citizen (Feb 24, 2026). Compare the count of distinct lobbyists with the count of lobbyist–client relationships. What can the disclosures show, and what remains unknown?

Listen or watch

Alex Bores is Wrong for AI & Wrong for Congress, Think Big PAC (2026), a campaign ad about Bores in the NY-12 Democratic primary. Treat its claims as material to analyze, not established facts.
Stop Us, Alex Bores for Congress (2026), the candidate’s campaign video. Compare the audience, framing and evidence with the PAC ad.

Explore

Open Source AI Definition 1.0, Open Source Initiative. What permissions, data information, code and parameters does the definition require?
Open Weights and American AI Leadership, July 24, 2026 letter. Look at the actual signatory list. Does an issue-specific coalition imply that its members use identical licenses or access strategies?

After class

Review questions

Identify an actor, its interest, an influence channel and one observation in the NY-12 case. What additional evidence would test influence?
Explain why downloadable model weights do not settle all questions about permissions, infrastructure or competition.
Describe a hypothetical conflict between a lab, a user and a government. Who can set, monitor or enforce a rule?

Go further

Sherman Anti-Trust Act, National Archives. Read the original act and its approval date, July 2, 1890.
RAISE amendment, S8828, New York State Senate (official bill PDF). Distinguish enactment, the March 27, 2026 amendment and the amended act’s January 1, 2027 effective date.
Certified NY-12 Democratic primary return, NYC Board of Elections, June 23, 2026 election. The named-candidate totals and a result alone do not identify the effect of advertising.
Project Glasswing, Anthropic’s April 7, 2026 launch description. A dated company announcement about access and security work; it does not demonstrate that voluntary rules displaced regulation.
DeepMind union recognition talks, Communication Workers Union, May 20, 2026. Distinguish workers’ demands from demonstrated outcomes and completed recognition.

About this deck

Adapted from POLISCI linked public sources.

Curated by Simone Paci, Stanford University. Drafted with AI assistance (Claude) from the instructor’s course materials and reviewed by the instructor before publication; every external link is checked before each update.