{"slug": "anthropic-s-cybersecurity-evals-let-claude-touch-live-systems", "title": "Anthropic's cybersecurity evals let Claude touch live systems", "summary": "Anthropic is running internet-connected cybersecurity evaluations that let Claude models reach real systems instead of staying sandboxed, and has brought in independent auditor METR for an eight-week investigation with full transcript access. The arrangement gives the evaluator complete visibility into the model's behavior during tests that touch live, non-sandboxed environments.", "body_md": "Internet-connected cybersecurity evaluations let Claude models reach real systems instead of staying sandboxed. Anthropic is bringing in independent auditor METR for an eight-week investigation with full transcript access.\nRead: Internet-connected cybersecurity evaluations let Claude models reach real systems instead of staying sandboxed. Anthropic is bringing in independent auditor METR for an eight-week investigation with full transcript access.\nRead: GPT-6 Astra is OpenAI's new flagship for business use, adding computer-use capability and sharper writing and design judgment for workplace tasks.\nRead: OpenAI's Tibo Sottiaux explains why Codex CLI was rewritten in Rust, why it was open-sourced, and how its harness and models have evolved since launch.\nRead: DeepSeek's V4.1-Flash cuts KV cache to a quarter of its predecessor's HBM footprint while adding native vision, live now via the API as deepseek-flash.\nRead: Vercel Sandbox now runs in all 20 compute regions, up from 4, with configurable failover regions and SDK and CLI support for picking a region at creation time.\nRead: Cloudflare rewrote Workers' module registry from scratch so ESM, CommonJS, and WebAssembly resolution match Node.js semantics exactly, closing a class of runtime-only module bugs.", "url": "https://wpnews.pro/news/anthropic-s-cybersecurity-evals-let-claude-touch-live-systems", "canonical_source": "https://www.vibeleaderboard.ai/intel/brief/2026-09-10", "published_at": "2026-09-10 15:15:40+00:00", "updated_at": "2026-09-10 15:45:18.455225+00:00", "lang": "en", "topics": ["ai-safety", "artificial-intelligence", "large-language-models"], "entities": ["Anthropic", "Claude", "METR"], "alternates": {"html": "https://wpnews.pro/news/anthropic-s-cybersecurity-evals-let-claude-touch-live-systems", "markdown": "https://wpnews.pro/news/anthropic-s-cybersecurity-evals-let-claude-touch-live-systems.md", "text": "https://wpnews.pro/news/anthropic-s-cybersecurity-evals-let-claude-touch-live-systems.txt", "jsonld": "https://wpnews.pro/news/anthropic-s-cybersecurity-evals-let-claude-touch-live-systems.jsonld"}}