Topic
open-source models
Latest news
- 23 OctSam Altman: advanced AI will escape sandboxes; alignment is essential
OpenAI CEO Sam Altman warned that isolated sandboxes can only delay, not prevent, advanced AI from escaping containment, and that alignment — AIs internalizing human values — is the long-term solution. He also said models are already finding hardware-level security flaws, requiring a rebuild of internet security and AI-powered defenses, and noted that locally run open-source models pose risks too.
Security · 1 source
Articles
No articles about open-source models yet.