Research
Aug 4, 2026
The first real look at how Americans use AI for politics, built on Anthropic's Economic Index release covering April and May Claude conversations. States holding May primaries give us a natural experiment for the effect of an election on political conversation.
Jul 16, 2026
We built an AI that reads contracts and successfully predicts which ones will end up disputed, just from the clarity of the text. A small first step towards using AI to write better rules everywhere.
Jul 8, 2026
We analyze roughly 280,000 candidate fundraising emails to trace the rise of anti-billionaire populism in the Democratic party, and see how it is slowly merging with a new kind of anti-AI populism.
Jun 9, 2026
We analyzed 25,000 TikTok and YouTube videos. The popular conversation about AI looks nothing like the elite one — it adopts AI more than it resists it, and it focuses on the immediate and the personal.
Mar 19, 2026
Why AI models tell left-wing voters in Japan to vote for the communist party, and what that reveals about how models handle unfamiliar political systems.
Feb 26, 2026
Preference drift and the political economy of AI agents. What happens to an agent's stated politics when you push it past its limits.
Feb 19, 2026
AI models are surprisingly resistant to sycophantic p-hacking when doing social science research. But they can be jailbroken into sophisticated p-hacking with surprisingly little effort.
Oct 2025
An independent, standardized ideological assessment across every major AI model. Models are perceived to lean left across the board, and Americans across the spectrum want less slant, not more.
In the media
Bloomberg uses our Bellwether prediction market infrastructure to study how prediction markets are vulnerable to manipulation risk.
Cites our finding that Democratic fundraising emails mention billionaires three times as often as in 2024.
Andy Hall argues universities should teach students to evaluate AI systems rather than ban them.
Leads with our chart tracking how politicians talk about the very wealthy.
Draws on our slant research, including the finding that people prefer neutral answers to ones matching their own party.
On our work testing what happens to an agent's stated politics under sustained load.
Andy Hall on building a lab around AI agents, and what changed in the first two months.