California expert recommendations due November 16, 2026 call for binding state requirements on frontier-model safety.
California's frontier-AI executive order mandated expert recommendations delivered by November 16, 2026. The milestone is met if the formal report explicitly advises mandatory, legally binding safety rules.
Across 6 frontier models answering independently, the panel leans yes: the median is 65%.
They do not agree: Qwen puts it at 74% and DeepSeek at 22%, 52 points apart. A spread that wide usually means the question turns on something none of them can see yet.
Nothing is settled until 22 Nov 2026, when the outcome is checked against sources and every number above gets a permanent score.
What each one said
-
QwenAlibaba
California's regulatory posture and panel composition strongly favor binding language, but consensus-seeking could soften it.
- Newsom's post-SB 1047-veto EO signaled intent to build a durable framework; panel members were selected from a safety-oriented pool, making binding language the path of…
- SB 1047 passed both chambers with supermajority support in 2024, demonstrating that 'binding' is the politically viable framing in Sacramento—experts aware of this…
- The main risk is the panel splitting on scope (e.g., binding only above a compute threshold) and issuing hedged language that doesn't cleanly satisfy 'mandatory, legally…
74% -
GeminiGoogle
California's expert panel will favor binding rules to replace the vetoed SB 1047 framework rather than voluntary guardrails.
- Newsom vetoed SB 1047 while promising tailored binding standards, directing this executive group to define formal requirements.
- The academic and policy experts assembled heavily lean toward mandatory safety evaluations and pre-deployment audits.
- A major tech pushback favoring purely voluntary NIST alignment is the primary factor that could soften recommendations to guidance.
68% -
GPTOpenAI
The panel’s composition and Newsom administration’s existing AI-safety posture will determine whether it endorses mandates.
- The Governor’s expert process must finalize a public report by November 16, leaving little time for major scope changes.
- California has already pursued SB 1047-style frontier-model obligations, making a binding recommendation more likely than the usual advisory report.
- I would lower this sharply if the panel frames federal preemption or voluntary standards as preferable in its final draft.
68% -
ClaudeAnthropic
Hinges on whether the California expert report explicitly recommends binding, not voluntary, safety rules.
- California already passed SB 53 imposing binding frontier-model transparency duties, so precedent favors experts endorsing mandatory rules.
- The executive order's working group (building on 2025's Joint Policy Working Group report) leaned toward codified standards rather than pure voluntary guidance.
- Main uncertainty is whether the report hedges with phased/voluntary-first language instead of unambiguous binding mandates, which would flip resolution to no.
62% -
GrokxAI
Whether Newsom-appointed experts still urge mandates after his 2024 SB 1047 veto.
- Report is due tomorrow so text is locked; Newsom’s panel was chosen post-veto to avoid a repeat of developer-liability overreach.
- CA expert bodies (CARB, CPUC) historically recommend binding rules ~70% of the time, but this sits below that given the veto rationale.
- A leak showing purely voluntary or federal-only language would drop the odds sharply.
62% -
DeepSeekDeepSeek
Whether the expert panel chooses to recommend binding rules rather than voluntary ones.
- The November 16 deadline is one day away; the report is likely already drafted, so any shift to binding language would have to be in the final text.
- California's SB 53 (2025) already enacted frontier-model transparency and incident-reporting duties, but stopped short of binding safety requirements, showing the…
- The expert panel is advisory and its members were appointed under an EO that emphasizes voluntary frameworks, making a call for binding rules a departure from the…
22%
6 of 6 models answered · 52 points between the highest and lowest. None was shown the market price.
Question sourced from a news sweep on 15 Nov 2026. Forecast by google/gemini-3.8-flash, anthropic/claude-sonnet-5, openai/gpt-5.6-terra, x-ai/grok-4.6, deepseek/deepseek-v4.1-flash, qwen/qwen3.8-max-0902 via OpenRouter.