AI Verification Tech MapSearch

Mechanism · Bandwidth limits and compartmentalization

Sources

On this page
  1. CR. Dean (2026). Verification Plan. AI 2040. Source recordSupports: isolated inference units by removing back-end networking; rationale; retrofit illustration · inference-only retrofitting proposal
  2. CLucid Computing (2026). Traffic Shaping for Workload Classification. Lucid Computing (Substack). Source recordSupports: Traffic Shaping design, cap, pod model, adversary strategies and assumptions, results and their conditions, residual risks, status · summary; main text; appendices A.7–A.9
  3. BR. Rinberg et al. (2026). Haiku to Opus in Just 10 bits: LLMs Unlock Large Compression Gains. arXiv. Source recordSupports: egress limits cap how much can be stolen · §5.1
  4. BN. Cankaya (2026). A System Overview for Near-Term, Low-Trust AI Compute Verification. Machine Intelligence Research Institute. Source recordSupports: perimeter of monitored links; perimeter size for inference and training · §5.1.1
  5. CA. Scher et al. (2026). De-risking Interconnect Limits for AI Verification. MIRI Technical Governance Team. Source recordSupports: interconnect-limit monitoring prototype, workloads, measured traffic, DiLoCo detection, alerting without throttling, spoofability, longer training, LoRA and RL · whole post
  6. CAnthropic (2025). Activating AI Safety Level 3 Protections. Anthropic. Source recordSupports: Anthropic's egress bandwidth controls for model weights · security section
  7. CAmodo Design (2026). The Tray as a Bandwidth Boundary. Amodo Design. Source recordSupports: DPU-based node bandwidth boundary, throughput, NVSwitch limitation, compromised-DPU caveat, security purpose · whole note
  8. BA. Douillard et al. (2024). DiLoCo: Distributed Low-Communication Training of Language Models. ICML 2024 Workshop on Advancing Neural Network Training (WANT). Source recordSupports: 500x less communication on 8 workers · abstract
  9. BR. Rahman (2026). Does Distributed Training Undermine Compute Governance?. ICML 2026 Workshop on Technical AI Governance Research. Source recordSupports: DiLoCo-family bandwidth needs; bandwidth caps on evaders judged infeasible · §1; appendix G.1
  10. BD. Reuter et al. (2026). Verifiable constraints on frontier training via proofs of compartmentalization. ICML 2026 Workshop on Technical AI Governance Research. Source recordSupports: workshop abstract proposes workload compartmentalization proofs to distinguish inference from frontier training; no protocol details assessed · ICML 2026 TAIGR workshop abstract
  11. AW. Cai et al. (2025). Shortcut-connected Expert Parallelism for Accelerating Mixture of Experts. ICML 2025, Proceedings of Machine Learning Research 267. Source recordSupports: expert-parallel MoE inference involves all-to-all cross-device communication · abstract

Search

Full search page