Palimpsest
Establishes a public, tamper-evident record of AI model evaluations and censorship actions, verifiable offline to ensure data integrity.
Advertisement
Establishes a public, tamper-evident record of AI model evaluations and censorship actions, verifiable offline to ensure data integrity.
Advertisement
Palimpsest acts as a crucial transparency primitive, providing a sealed, append-only ledger that allows anyone to verify, offline, that data has not been altered after publication. It serves two primary functions: a Verifiable Eval Registry for AI models, where evaluation results are sealed at publication with pre-committed probe sets to prevent post-hoc changes, and a Censorship Observatory that quantifies authoritarian deletion and content-layer censorship using public data sources. This unique architecture ensures the integrity of sensitive information, offering an auditable record of what powerful entities quietly erase, without requiring trust in any central authority.