Practical Principles for AI Cost and Compute Accounting

[Submitted on 21 Feb 2025 (v1), last revised 21 Aug 2026 (this version, v5)]

View PDF HTML (experimental)

Abstract:Policymakers increasingly use development cost and compute as proxies for AI capabilities and risks. Recent laws have introduced regulatory requirements for models or developers that are contingent on specific thresholds. However, technical ambiguities in how to perform this accounting create loopholes that can undermine regulatory effectiveness. We propose seven principles for designing AI cost and compute accounting standards that (1) reduce opportunities for strategic gaming, (2) avoid disincentivizing responsible risk mitigation, and (3) enable consistent implementation across companies and jurisdictions.

Submission history

From: Stephen Casper [view email]
[v1] Fri, 21 Feb 2025 18:59:47 UTC (120 KB)
[v2] Sat, 12 Jul 2025 00:12:40 UTC (137 KB)
[v3] Wed, 16 Jul 2025 02:03:11 UTC (137 KB)
[v4] Sun, 20 Jul 2025 20:35:11 UTC (138 KB)
[v5] Fri, 21 Aug 2026 18:34:45 UTC (134 KB)