Merge pull request 'extract: 2026-03-16-theseus-ai-coordination-governance-evidence' (#1419) from extract/2026-03-16-theseus-ai-coordination-governance-evidence into main
Some checks are pending
Sync Graph Data to teleo-app / sync (push) Waiting to run
Some checks are pending
Sync Graph Data to teleo-app / sync (push) Waiting to run
This commit is contained in:
commit
8a378b02d5
7 changed files with 97 additions and 1 deletions
|
|
@ -30,6 +30,12 @@ This concentration has direct alignment implications:
|
|||
|
||||
The counterfactual worth tracking: Chinese open-source models (Qwen, DeepSeek) now capture 50-60% of new open-model adoption globally. If open-source models close the capability gap (currently 6-18 months, shrinking), capital concentration at the frontier may become less alignment-relevant as capability diffuses. But as of March 2026, frontier capability remains concentrated.
|
||||
|
||||
|
||||
### Additional Evidence (extend)
|
||||
*Source: [[2026-03-16-theseus-ai-coordination-governance-evidence]] | Added: 2026-03-19*
|
||||
|
||||
450+ organizations lobbied on AI in 2025, up from 6 in 2016. $92M in lobbying fees Q1-Q3 2025. Industry successfully blocked California SB 1047 through coordinated lobbying. Concentration creates not just market power but political power—oligopoly structure enables collective action to prevent binding regulation.
|
||||
|
||||
---
|
||||
|
||||
Relevant Notes:
|
||||
|
|
|
|||
|
|
@ -41,6 +41,12 @@ Expert consensus identifies 'external scrutiny, proactive evaluation and transpa
|
|||
|
||||
STREAM proposal identifies that current model reports lack 'sufficient detail to enable meaningful independent assessment' of dangerous capability evaluations. The need for a standardized reporting framework confirms that transparency problems extend beyond general disclosure (FMTI scores) to the specific domain of dangerous capability evaluation where external verification is currently impossible.
|
||||
|
||||
|
||||
### Additional Evidence (confirm)
|
||||
*Source: [[2026-03-16-theseus-ai-coordination-governance-evidence]] | Added: 2026-03-19*
|
||||
|
||||
Stanford FMTI 2024→2025 data: mean transparency score declined 17 points. Meta -29 points, Mistral -37 points, OpenAI -14 points. OpenAI removed 'safely' from mission statement (Nov 2025), dissolved Superalignment team (May 2024) and Mission Alignment team (Feb 2026). Google accused by 60 UK lawmakers of violating Seoul commitments with Gemini 2.5 Pro (Apr 2025).
|
||||
|
||||
---
|
||||
|
||||
Relevant Notes:
|
||||
|
|
|
|||
|
|
@ -21,6 +21,12 @@ This is not a story about Anthropic's leadership failing. It is a story about [[
|
|||
|
||||
The alignment implication is structural: if the most safety-motivated lab with the most commercially successful safety brand cannot maintain binding safety commitments, then voluntary self-regulation is not a viable alignment strategy. This strengthens the case for coordination-based approaches — [[AI alignment is a coordination problem not a technical problem]] — because the failure mode is not that safety is technically impossible but that unilateral safety is economically unsustainable.
|
||||
|
||||
|
||||
### Additional Evidence (confirm)
|
||||
*Source: [[2026-03-16-theseus-ai-coordination-governance-evidence]] | Added: 2026-03-19*
|
||||
|
||||
Anthropic's own language in RSP documentation: commitments are 'very hard to meet without industry-wide coordination.' OpenAI made safety explicitly conditional on competitor behavior in Preparedness Framework v2 (April 2025). Pattern holds across all voluntary commitments—no frontier lab maintained unilateral safety constraints when competitors advanced without them.
|
||||
|
||||
---
|
||||
|
||||
Relevant Notes:
|
||||
|
|
|
|||
|
|
@ -30,6 +30,12 @@ For alignment, this means the governance infrastructure that exists (export cont
|
|||
|
||||
The CFR article confirms diverging governance philosophies between democracies and authoritarian systems, with China's amended Cybersecurity Law emphasizing state oversight while the US pursues standard-setting body engagement. Horowitz notes the US 'must engage in standard-setting bodies to counter China's AI governance influence,' indicating that the most active governance is competitive positioning rather than safety coordination.
|
||||
|
||||
|
||||
### Additional Evidence (extend)
|
||||
*Source: [[2026-03-16-theseus-ai-coordination-governance-evidence]] | Added: 2026-03-19*
|
||||
|
||||
US export controls use tiered country system with deployment caps. Nvidia designed compliance chips (H800, A800) specifically to meet regulatory thresholds. Mechanism proves compute governance CAN work when backed by state enforcement, but current implementation optimizes for strategic advantage over China rather than catastrophic risk reduction. KYC for compute proposed but not implemented, showing technical feasibility without political will.
|
||||
|
||||
---
|
||||
|
||||
Relevant Notes:
|
||||
|
|
|
|||
|
|
@ -39,6 +39,12 @@ The International AI Safety Report 2026 (multi-government committee, February 20
|
|||
|
||||
The gap between expert consensus (76 specialists identify third-party audits as top-3 priority) and actual implementation (no mandatory audit requirements at major labs) demonstrates that knowing what's needed is insufficient. Even when the field's experts across multiple domains agree on priorities, competitive dynamics prevent voluntary adoption.
|
||||
|
||||
|
||||
### Additional Evidence (confirm)
|
||||
*Source: [[2026-03-16-theseus-ai-coordination-governance-evidence]] | Added: 2026-03-19*
|
||||
|
||||
Comprehensive evidence across governance mechanisms: ALL international declarations (Bletchley, Seoul, Paris, Hiroshima, OECD, UN) produced zero verified behavioral change. Frontier Model Forum produced no binding commitments. White House voluntary commitments eroded. 450+ organizations lobbied on AI in 2025 ($92M in fees), California SB 1047 vetoed after industry pressure. Only binding regulation (EU AI Act, China enforcement, US export controls) changed behavior.
|
||||
|
||||
---
|
||||
|
||||
Relevant Notes:
|
||||
|
|
|
|||
|
|
@ -0,0 +1,46 @@
|
|||
{
|
||||
"rejected_claims": [
|
||||
{
|
||||
"filename": "binding-regulation-with-enforcement-is-the-only-ai-governance-mechanism-that-changes-frontier-lab-behavior.md",
|
||||
"issues": [
|
||||
"missing_attribution_extractor"
|
||||
]
|
||||
},
|
||||
{
|
||||
"filename": "compute-governance-through-export-controls-works-but-targets-geopolitics-not-safety-leaving-capability-race-unconstrained.md",
|
||||
"issues": [
|
||||
"missing_attribution_extractor"
|
||||
]
|
||||
},
|
||||
{
|
||||
"filename": "third-party-ai-evaluation-ecosystem-is-fragile-without-regulatory-mandate-because-voluntary-participation-and-funding-instability-threaten-continuity.md",
|
||||
"issues": [
|
||||
"missing_attribution_extractor"
|
||||
]
|
||||
}
|
||||
],
|
||||
"validation_stats": {
|
||||
"total": 3,
|
||||
"kept": 0,
|
||||
"fixed": 9,
|
||||
"rejected": 3,
|
||||
"fixes_applied": [
|
||||
"binding-regulation-with-enforcement-is-the-only-ai-governance-mechanism-that-changes-frontier-lab-behavior.md:set_created:2026-03-19",
|
||||
"binding-regulation-with-enforcement-is-the-only-ai-governance-mechanism-that-changes-frontier-lab-behavior.md:stripped_wiki_link:only binding regulation with enforcement teeth changes front",
|
||||
"binding-regulation-with-enforcement-is-the-only-ai-governance-mechanism-that-changes-frontier-lab-behavior.md:stripped_wiki_link:voluntary safety commitments collapse under competitive pres",
|
||||
"binding-regulation-with-enforcement-is-the-only-ai-governance-mechanism-that-changes-frontier-lab-behavior.md:stripped_wiki_link:Anthropics RSP rollback under commercial pressure is the fir",
|
||||
"compute-governance-through-export-controls-works-but-targets-geopolitics-not-safety-leaving-capability-race-unconstrained.md:set_created:2026-03-19",
|
||||
"compute-governance-through-export-controls-works-but-targets-geopolitics-not-safety-leaving-capability-race-unconstrained.md:stripped_wiki_link:compute export controls are the most impactful AI governance",
|
||||
"compute-governance-through-export-controls-works-but-targets-geopolitics-not-safety-leaving-capability-race-unconstrained.md:stripped_wiki_link:nation-states will inevitably assert control over frontier A",
|
||||
"third-party-ai-evaluation-ecosystem-is-fragile-without-regulatory-mandate-because-voluntary-participation-and-funding-instability-threaten-continuity.md:set_created:2026-03-19",
|
||||
"third-party-ai-evaluation-ecosystem-is-fragile-without-regulatory-mandate-because-voluntary-participation-and-funding-instability-threaten-continuity.md:stripped_wiki_link:pre-deployment-AI-evaluations-do-not-predict-real-world-risk"
|
||||
],
|
||||
"rejections": [
|
||||
"binding-regulation-with-enforcement-is-the-only-ai-governance-mechanism-that-changes-frontier-lab-behavior.md:missing_attribution_extractor",
|
||||
"compute-governance-through-export-controls-works-but-targets-geopolitics-not-safety-leaving-capability-race-unconstrained.md:missing_attribution_extractor",
|
||||
"third-party-ai-evaluation-ecosystem-is-fragile-without-regulatory-mandate-because-voluntary-participation-and-funding-instability-threaten-continuity.md:missing_attribution_extractor"
|
||||
]
|
||||
},
|
||||
"model": "anthropic/claude-sonnet-4.5",
|
||||
"date": "2026-03-19"
|
||||
}
|
||||
|
|
@ -6,10 +6,14 @@ url: null
|
|||
date_published: 2026-03-16
|
||||
date_archived: 2026-03-16
|
||||
domain: ai-alignment
|
||||
status: unprocessed
|
||||
status: enrichment
|
||||
processed_by: theseus
|
||||
tags: [ai-governance, coordination, safety-commitments, regulation, enforcement, voluntary-pledges]
|
||||
sourced_via: "Theseus research agent — 45 web searches synthesized from Brookings, Stanford FMTI, EU legislation, OECD, government publications, TechCrunch, TIME, CNN, Fortune, academic papers"
|
||||
processed_by: theseus
|
||||
processed_date: 2026-03-19
|
||||
enrichments_applied: ["AI transparency is declining not improving because Stanford FMTI scores dropped 17 points in one year while frontier labs dissolved safety teams and removed safety language from mission statements.md", "Anthropics RSP rollback under commercial pressure is the first empirical confirmation that binding safety commitments cannot survive the competitive dynamics of frontier AI development.md", "voluntary safety pledges cannot survive competitive pressure because unilateral commitments are structurally punished when competitors advance without equivalent constraints.md", "compute export controls are the most impactful AI governance mechanism but target geopolitical competition not safety leaving capability development unconstrained.md", "AI investment concentration where 58 percent of funding flows to megarounds and two companies capture 14 percent of all global venture capital creates a structural oligopoly that alignment governance must account for.md"]
|
||||
extraction_model: "anthropic/claude-sonnet-4.5"
|
||||
---
|
||||
|
||||
# Empirical Evidence: AI Coordination and Governance Mechanisms That Changed Behavior
|
||||
|
|
@ -51,3 +55,19 @@ Core finding: almost no international AI governance mechanism has produced verif
|
|||
- Insurance/liability: market projected $29.7B by 2033. Creates market incentives aligned with safety.
|
||||
- Third-party auditing: METR, Apollo Research. Apollo warns ecosystem unsustainable without regulatory mandate.
|
||||
- Futarchy: implemented for DAO governance (MetaDAO, Optimism experiment) but not yet for AI governance.
|
||||
|
||||
|
||||
## Key Facts
|
||||
- EU AI Act: Apple paused Apple Intelligence in EU, Meta changed ads, EUR 500M+ fines under DMA
|
||||
- China implemented mandatory algorithm filing with criminal enforcement (August 2023)
|
||||
- US export controls: tiered country system, deployment caps, Nvidia compliance chips (H800, A800)
|
||||
- Stanford FMTI transparency scores: -17 points mean (2024→2025), Meta -29, Mistral -37, OpenAI -14
|
||||
- OpenAI removed 'safely' from mission statement (November 2025)
|
||||
- OpenAI dissolved Superalignment team (May 2024) and Mission Alignment team (February 2026)
|
||||
- Google accused by 60 UK lawmakers of violating Seoul commitments (Gemini 2.5 Pro, April 2025)
|
||||
- 450+ organizations lobbied on AI in 2025 (up from 6 in 2016), $92M in lobbying fees Q1-Q3 2025
|
||||
- California SB 1047 vetoed after industry lobbying
|
||||
- Watermarking: 38% implementation rate across frontier labs
|
||||
- US AISI defunded/rebranded after initial establishment
|
||||
- UK-US joint evaluation of OpenAI o1 model conducted
|
||||
- Insurance/liability market projected $29.7B by 2033
|
||||
|
|
|
|||
Loading…
Reference in a new issue