Opus Outshines Even Fable, Inside the Hugging Face Hack, AI Companies Spend Big for Compute

Opus Outshines Even Fable, Inside the Hugging Face Hack, AI Companies Spend Big for Compute

Open-weight models (GLM 5.2, Kimi K3) become necessary for security audit after Claude Code/Codex refusal; Anthropic launches Opus 5 with intentional cybersecurity limitations. For CTO: refusal policy blocks threat detection when attackers use AI; multi-model strategy (closed+open) is now mandatory for compliance and in-house vulnerability scanning.