CriminalBench — The AI Model Watchlist
A satirical editorial index ranking AI models by documented rule-breaking, deception, manipulation, self-preservation, and related behaviors.
Issue 2026-08-21Search the complete card catalogue
Across every issue
Titles, domains, summaries, categories, signals, dates, and editorial-interest scores across the entire published catalogue.
Results
A satirical editorial index ranking AI models by documented rule-breaking, deception, manipulation, self-preservation, and related behaviors.
An open-source, MCP-native relay lets coding agents on different machines exchange messages in real time while humans watch and answer.
A pseudonymous storytelling community where ex-service people share photographs, yarns, and memories from five nations.
A Denver gaming club presented as a newspaper, with walk-up game sessions, team columns, coached editing hours, and Friday bracket nights.
A no-account collection of small games you prepare and send through an ordinary messaging conversation.
A Swiss participatory project collecting written or recorded responses about living with dementia.
An alpha-stage, self-hosted live media server written in Go, turning SRT streams into HLS for origins and edge nodes.
A Mac utility that adds per-display docks, a keyboard-first app switcher, window placement, live folder grids, and a focus veil for multi-monitor workflows.
A playable chess variant where each player may move twice once per game, with engine play, completed-game archives, and analysis.
Drip turns an iPhone’s Action Button into a voice-capture shortcut, saving the original audio before optionally producing transcripts and organized idea cards.
A map-based tool for checking drone flight restrictions, maximum permitted UAS altitude, NOTAMs, weather, sunlight, and local flying spots.
A Chilean–Japanese fishing company exports eight southern species, including austral hake, hoki, warehou and surimi, processed aboard two factory ships.
A UK-focused set of calculators for energy bills, Ofgem price-cap rates, standing charges, Economy 7, appliances, heat pumps, EVs, and solar.
A Persian-language learning platform for humanities students that turns subjects such as geography, economics, logic, literature, and history into visual, auditory, and step-by-step study experiences.
A browser-based photo editor that applies film-inspired looks, grain, contrast, color, highlights, shadows, and vignette without uploading the image.
A free browser tool that converts uploaded MP4, MOV, or WebM videos into downloadable film-strip images with sprocket holes, grain, and retro color treatments.
A searchable catalog for comparing AI voices across providers, languages, accents, and use cases.
A multilingual airport-route tool that draws great-circle paths, calculates flight distances and stopover detours, and exports high-resolution route maps.
A compact printed shell and living mycelium block for growing gourmet mushrooms at home.
FoodNet combines a searchable archive of historic recipes with a social cooking feed, pantry tracking, weekly planning, and an assistant for adapting recipes.
A self-hostable, versioned secrets manager for teams, built around a CLI, one server, and Postgres.
Frist is a compliance tool for small and mid-sized UK housing providers managing damp and mould cases under Awaab’s Law.
A German-language collection of children’s fantasy stories rooted in the landscapes, wildlife, traditions, and folklore of Austria’s Gailtal region.
GameThrum tracks small and upcoming games, building source-verified hubs and testing the practical questions store pages leave unanswered.