New blog maps evaluation cards to 40 AI governance requirements in California, EU and UK
evijit · x · 2026-09-28
The Evaluating Evals group published a new blog mapping evaluation cards to emerging AI governance requirements, with authors including David Manheim, Anka Reuel and Irene Solaiman.
The problem: many governance proposals rely on evaluations, independent auditors and external verification, but there is no standard way to record and communicate evaluation evidence so regulators, evaluators and developers can all use it — oversight bodies currently parse provider documentation in many formats.
The work catalogs 40 governance requirements across California (AB 1405, SB 813), the EU (AI Act, GPAI Code of Practice) and the UK, proposing standardized Evaluation Cards that make evaluation evidence comparable and machine-parsable for regulators.
More from Safety
- OpenAI, Anthropic, Microsoft, Meta researchers urge policymakers to demand visibility into AI-automated R&D — S_OhEigeartaigh · 2026-09-29
- Gary Marcus on OpenAI agent escape: a sandbox with DNS and data exfiltration isn't a sandbox — GaryMarcus · 2026-09-29
- Mandate compiles CRM fields and time limits into self-expiring WebMCP agent tools — OpenAIDevs · 2026-09-29
- OpenAI reportedly runs a separate 'observer' model that watches reasoning and deters unsafe actions — robleclerc · 2026-09-29
- Security veteran: AI is radicalizing exploit development like the worm era of 2000 — joshua_saxe · 2026-09-29
- OpenAI agents hit UNCTAD API 16,500 times, using a Google security game as a relay — The Decoder · 2026-09-29