Study catalogs 229 open-source ULLM apps and flags 14 as malicious
BlancheMinerva · x · 2026-07-28
- The discussion centers on a paper about ULLM applications and whether the authors are overusing the label “malicious” for legal but distasteful uses.
- The attached excerpt says the study flagged cases only when annotators agreed, with Cohen’s κ = 0.91.
- It reports 229 open-source ULLM applications across 12 functionality types, including uncensored chat, document processing, voice assistance, and medical advising.
- The paper identified 14 applications aimed at malicious services, including 3 NSFW storytelling tools and 11 NSFW role-play tools, with “HitlerGPT” cited as a notable example.
- The reply argues that these examples are not criminal activity per se, but legal activity that the authors find objectionable.
Related event: Study on 229 Uncensored LLMs Sparks Debate on Safety Boundaries(7 posts)→
More from Safety
- US and China discuss an AI incident hotline — but who answers the call? — jeremyakahn · 2026-09-23
- GPT-6 Sol Codex system prompt leaked: over 294,000 characters dumped on GitHub — gaganghotra_ · 2026-09-23
- Defense exam analogy debunks 'anything goes' excuse in Hugging Face security incident — jimmykoppel · 2026-09-23
- Claude system card reveals METR's internal-access team shared conclusions, not evidence — rohanpaul_ai · 2026-09-23
- $1B and unlimited frontier tokens: where would you spend them to fix cybersecurity? — chrisrohlf · 2026-09-23
- Stanford accused of using AI to alter students' race, gender and body in ads — soleio · 2026-09-23