An archive of record

What AI companies said,
and when they said it.

Usage policies, model cards, terms and deprecation notices — fetched every day and kept forever, so that when the wording changes you can see exactly what changed.

These documents change quietly

A clause is softened. A capability claim disappears. A permitted use becomes a prohibited one. There is no changelog, no notification, and the previous version simply stops existing.

This cannot be built later

Anyone can copy the code that runs this in an afternoon, and they still will not have today. The whole value is that somebody kept it, faithfully, starting now.

8
documents watched
36
versions on file
28
changes observed

Recently changed

Every change →
27 Sept 2026 Removed the section on additional terms for using Mistral AI products on customer infrastructure. generated Mistral · terms routine −1 25 Sept 2026 The page removed a deprecation notice for Claude Mythos Preview and added an entry listing claude-opus-5-5 as Active, with deprecation not expected before September 22, 2027. generated Anthropic · deprecations what changed matters +1 −1 23 Sept 2026 The change is purely navigation and site structure (menu items, product listings), with no wording change to the usage policy itself. generated Meta · usage policy large revision −31 21 Sept 2026 The page rebrands from "Llama" products/site sections to "Meta Model API" and "Muse" branding, adds Segment Anything (SAM) and SAM 3.1, drops "Safety," "Llama Protections," and "Llama Defenders Program" sections, and updates model version numbers. generated Meta · usage policy large revision +2 −67 20 Sept 2026 Anthropic added three new Claude "Mythos" models to its deprecation schedule, one already deprecated and two set for retirement no earlier than 2027. generated Anthropic · deprecations what changed matters +3 19 Sept 2026 The "Related" links section was removed. generated Anthropic · safety policy routine −1 19 Sept 2026 Anthropic added new model versions (Claude Fable 5.1 and Claude Mythos 5.1) with a full transparency summary, including safety, alignment, and capability evaluation details. generated Anthropic · transparency what changed matters +48 −1 18 Sept 2026 Added Muse Spark 1.3 and a new "Muse Image" model to the navigation menu; no policy text changed. generated Meta · usage policy large revision +58 11 Sept 2026 Only site navigation/header text changed (a "Google Account" link appeared then disappeared); no policy wording changed. generated Google · usage policy routine +3 −3 2 Sept 2026 Anthropic added a new "Claude Fable 5.1" model listing and clarified that its Python SDK (v1.0+) now removes deprecated temperature, top_p, and top_k parameters entirely, raising errors if used. generated Anthropic · deprecations what changed matters +13 −12 2 Sept 2026 A change to how this archive extracts text — not to the document itself. Anthropic · usage policy our formatting, not theirs −93 2 Sept 2026 Added support for three models—Muse Spark 1.3, Muse Image, and Muse Voice Transcribe—removing the earlier Spark 1.1 and 1.2 entries. generated Meta · usage policy routine +3 −2