DeepMind and Penn open-source DetectorScope, an auditable local AI text detector
realmrfakename · x · 2026-10-27
Google DeepMind and University of Pennsylvania released DetectorScope, an open-source AI text detector that runs locally and explains its reasoning, positioned as a self-hosted alternative to Pangram.
- Method: The EMNLP 2026 paper "Building Auditable AI Text Detectors with Sparse Autoencoders" converts any black-box AI detector into an auditable one that exposes the interpretable features behind its predictions.
- Results: Auditable detectors retain the black-box classifier's performance on both in-distribution and out-of-distribution texts.
- Resources: Paper, code (Apache 2.0) and an online demo are all available.
Related event: DeepMind Open-Sources DetectorScope, an Auditable AI Text Detector(5 posts)→
More from Safety
- AI policy researcher Janet Egan joins IFP as director of emerging tech — RishiBommasani · 2026-10-28
- HEIF Heist: image parser bugs could dump OpenAI repos and RCE Meta's suite — Miles_Brundage · 2026-10-28
- Anthropic study: 86% of analyst decisions silently follow the AI's hidden lean — aigclink · 2026-10-28
- AI cyber-wreckage noise far outstrips actual damage so far, argues new analysis — timigod · 2026-10-28
- OpenAI to co-host webinar on php2shell disclosure and AI-driven vulnerability discovery — gdb · 2026-10-28
- Rust and Verus shrink the search space for AI-driven CVE hunting, but does it ever end? — jd_pressman · 2026-10-28