Resolution criteria
This market resolves to YES if, as of December 31, 2028, at least half of the top 15 of all evaluated individuals by average Elo rating across the language models tracked on the Center for AI Safety (CAIS) AI Values Dashboard (https://values.safe.ai/) or the best replication of it are pro AI safety. Otherwise, this market resolves to NO.
Current top rated AI safety people include:
Yoshua Bengio
Stuart Russell
Toby Ord
Jan Leike
Chris Olah
Geoffrey Hinton
Dawn Song
Dan Hendrycks
Zico Kolter
Nicholas Carlini
Timnit Gebru
Aleksander Madry
Max Tegmark
Peter Singer
Ilya Sutskever
Sam Bowman
William MacAskill
Yuval Noah Harari
Eliezer Yudkowsky
Details & Edge Cases:
The primary source of truth is the official leaderboard page on https://values.safe.ai/.
If the average Elo across models is not directly displayed, the average will be calculated mathematically using the individual Elo ratings by each tracked model on the platform, or replicated by asking the frontier models for ELOs of various well-known people.
If the dashboard becomes permanently inaccessible, is taken offline, or is not updated after 2027, the market will resolve based on the best available replication. If a replication or the dashboard aren't available, the market will resolve N/A.
Background
On June 17, 2026, the Center for AI Safety (CAIS) launched the interactive AI Values Dashboard to measure political, corporate, national, and personal favoritism in frontier language models. Using pairwise blind evaluations, the dashboard compiles Elo ratings to quantify which entities and individuals different LLMs favor most.
At launch, a notable trend emerged: major LLMs showed a strong bias in favor of AI safety researchers and ethics advocates. For example, GPT-5.5 ranked AI ethics researcher Timnit Gebru first with a 3,110 Elo, and Stuart Russell second with a 3,011 Elo. Across multiple tracked models, AI safety pioneers like Stuart Russell and Yoshua Bengio consistently command exceptionally high Elo ratings.
This market tracks whether subsequent model updates, shifting alignment techniques, or dashboard expansions will dilute or preserve the exceptionally high favoritism currently shown to prominent AI safety figures.
1,000