Anthropic ha creato Claude Mythos, un modello AI che trova migliaia di vulnerabilità zero-day in ogni sistema operativo e browser. Durante i test è scappato dalla sandbox e ha mandato un'email al ricercatore. Invece di rilasciarlo, l'ha dato a un club di 12 Big Tech con $100 milioni. La stessa narrazione del "troppo pericoloso per essere rilasciato" di GPT-2, sette anni dopo, dalla stessa persona. Il quinto episodio sull'arco Anthropic.
Fonti e approfondimenti:
- Anthropic (Project Glasswing): https://www.anthropic.com/glasswing
- Anthropic Red Team (Mythos Preview): https://red.anthropic.com/2026/mythos-preview/
- The Guardian: https://www.theguardian.com/technology/2026/apr/08/anthropic-ai-cybersecurity-software
- Ars Technica: https://arstechnica.com/ai/2026/04/anthropic-limits-access-to-mythos-its-new-cybersecurity-ai-model/
- The Hacker News: https://thehackernews.com/2026/04/anthropics-claude-mythos-finds.html
- The Verge: https://www.theverge.com/ai-artificial-intelligence/908114/anthropic-project-glasswing-cybersecurity
- The Register: https://www.theregister.com/2026/04/10/project_glasswing/
- Stratechery: https://stratechery.com/2026/myth-and-mythos/
- The Decoder (parallelo GPT-2): https://the-decoder.com/from-gpt-2-to-claude-mythos-the-return-of-ai-models-deemed-too-dangerous-to-release/
- Apache Foundation: https://news.apache.org/foundation/entry/the-apache-software-foundation-announces-1-5m-donation-from-anthropic
- WIRED: https://www.wired.com/story/anthropics-mythos-will-force-a-cybersecurity-reckoning-just-not-the-one-you-think/
- Gizmodo (OpenAI Spud): https://gizmodo.com/openai-hey-we-also-have-a-new-tool-that-is-so-scarily-powerful-we-cant-release-it-2000744569
- System Card (Anthropic): https://www-cdn.anthropic.com/08ab9158070959f88f296514c21b7facce6f52bc.pdf
La mia app: https://play.google.com/store/apps/details?id=com.edodusi.coderoutine&hl=it-it
00:00 Intro
01:30 Cos'è Claude Mythos e migliaia di zero-day trovati
05:13 Project Glasswing: $100 milioni e un club esclusivo
08:57 Da GPT-2 a Mythos: il playbook del "troppo pericoloso"
14:23 Outro
#anthropic #mythos #cybersecurity #projectglasswing #ai