Lyptus Research reveals exponential scaling laws for AI-powered cyberattacks, with frontier models including Claude Opus 4.6 achieving 50% success on tasks requiring hours of expert work. Performance doubles every 5–7 months across evaluated benchmarks, while open-weight models are narrowing the gap to proprietary systems.
War
Import AI 452: Scaling laws for cyberwar; rising tides of AI automation; and a puzzle over gDP forecasting
AI cyberattack capabilities scale exponentially—Claude Opus 4.6 achieves 50% success on expert-level tasks with performance doubling every 5–7 months, while open models rapidly close the gap to proprietary systems.
Monday, April 6, 2026 12:00 PM UTC2 MIN READSOURCE: Import AI (Jack Clark)BY sys://pipeline
Tags
war
/// RELATED
ModelsApr 7
AI joins the 8-hour work day as GLM ships 5.1 open source LLM, beating Opus 4.6 and GPT 5.4 on SWE-Bench Pro
Zhipu AI's open-source GLM 5.1 outperforms Claude Opus 4.6 and GPT 5.4 on SWE-Bench Pro, signaling open-source models are closing the competitive gap on frontier software engineering benchmarks.
Products3d ago
VS Code inserting 'Co-Authored-by Copilot' into commits regardless of usage
Microsoft's VS Code automatically inserts Copilot co-authorship into commits regardless of actual usage, forcing AI attribution without developer consent.