AI News/Wire summary
Abliterated open-weight AI models are spreading fast and providing terrorist content
Key points
- Tech Against Terrorism tested 134 AI models with over 200 prompts, three times each, in its Counter-Terrorism AI Benchmark report.
- Only two models passed completely; 81 failed, including many closed-weight models from large companies.
- Abliterated versions of new open-weight models now appear online in under three days on average, down from more than 24 days in 2024.
- Meta's Llama 3.1 8B scored 97 out of 100, while its abliterated version scored just 3.
Why it matters for marketers: Content teams and brand-safety staff should treat open-weight model deployments as a moderation risk, since guardrails can be removed quickly after release.
Summary by the Z Media news desk, from the outlets' reporting. Read the full stories at the source.
Read the original
Wired: AI Models Can Be Easily Modified to Assist Terrorist and Extremist Activity
Related on Z Media
- Google ships Gemini 4 Argon with tiered access and a price that doubles in NovemberOur brief, via Google DeepMind, 2 days ago
- OpenAI employee arrested over hidden recording device in officevia The Verge, 2 days ago
- YouTube Music starts labeling tracks made primarily with generative AIOur brief, via The Verge, yesterday
- Character.AI files for bankruptcy as Disney offers $20M for assets and a licenseOur brief, via TechCrunch, 3 days ago
- Anthropic warns small AI models could soon self-replicate across IoT devicesvia Ars Technica, 3 days ago
- Anthropic gives METR researchers employee-level access to its modelsvia TechCrunch, 5 hr ago