teleo/teleo-codex
8
0
Fork 0
Code Issues 22 Pull requests 41 Projects Releases Packages Wiki Activity Actions 1

extract: 2026-03-20-leo-nuclear-ai-governance-observability-gap

Browse source 
Pentagon-Agent: Epimetheus <3D35839A-7722-4740-B93D-51157F7D5E70>
This commit is contained in:
Teleo Agents 2026-03-20 08:16:42 +00:00
parent 8b309972a8
commit 3b933f6386
2 changed files with 48 additions and 1 deletions
Show stats Download patch file Download diff file Expand all files Collapse all files

34
inbox/queue/.extraction-debug/2026-03-20-leo-nuclear-ai-governance-observability-gap.json Normal file
View file

@ -0,0 +1,34 @@
{
"rejected_claims": [
{
"filename": "nuclear-governance-succeeded-through-physical-observability-not-political-will-making-the-template-inapplicable-to-ai.md",
"issues": [
"missing_attribution_extractor"
]
},
{
"filename": "ai-governance-mechanisms-regulating-physically-observable-inputs-are-structurally-more-durable-than-capability-evaluation.md",
"issues": [
"missing_attribution_extractor"
]
}
],
"validation_stats": {
"total": 2,
"kept": 0,
"fixed": 4,
"rejected": 2,
"fixes_applied": [
"nuclear-governance-succeeded-through-physical-observability-not-political-will-making-the-template-inapplicable-to-ai.md:set_created:2026-03-20",
"nuclear-governance-succeeded-through-physical-observability-not-political-will-making-the-template-inapplicable-to-ai.md:stripped_wiki_link:technology-advances-exponentially-but-coordination-mechanism",
"ai-governance-mechanisms-regulating-physically-observable-inputs-are-structurally-more-durable-than-capability-evaluation.md:set_created:2026-03-20",
"ai-governance-mechanisms-regulating-physically-observable-inputs-are-structurally-more-durable-than-capability-evaluation.md:stripped_wiki_link:technology-advances-exponentially-but-coordination-mechanism"
],
"rejections": [
"nuclear-governance-succeeded-through-physical-observability-not-political-will-making-the-template-inapplicable-to-ai.md:missing_attribution_extractor",
"ai-governance-mechanisms-regulating-physically-observable-inputs-are-structurally-more-durable-than-capability-evaluation.md:missing_attribution_extractor"
]
},
"model": "anthropic/claude-sonnet-4.5",
"date": "2026-03-20"
}

15
inbox/queue/2026-03-20-leo-nuclear-ai-governance-observability-gap.md
View file

@ -7,7 +7,7 @@ date: 2026-03-20
domain: grand-strategy
secondary_domains: [ai-alignment]
format: synthesis
status: unprocessed
status: null-result
priority: high
tags: [nuclear-analogy, observability-gap, AI-governance, physical-constraints, export-controls, grand-strategy, historical-analogy]
synthesizes:
@ -15,6 +15,10 @@ synthesizes:
- 2026-03-20-bench2cop-benchmarks-insufficient-compliance.md
- 2026-03-20-euaiact-article92-compulsory-evaluation-powers.md
- 2026-00-00-darioamodei-adolescence-of-technology.md
processed_by: leo
processed_date: 2026-03-20
extraction_model: "anthropic/claude-sonnet-4.5"
extraction_notes: "LLM returned 2 claims, 2 rejected by validator"
---
## Content
@ -81,3 +85,12 @@ The nuclear timeline (~23 years from Hiroshima to NPT) is often cited as evidenc
PRIMARY CONNECTION: [[technology advances exponentially but coordination mechanisms evolve linearly creating a widening gap]]
WHY ARCHIVED: Provides historical grounding for why the tech-governance gap is structural for AI (not just slow), and identifies the specific mechanism (observability) that makes nuclear governance work but AI governance fail
EXTRACTION HINT: Focus on the observability mechanism, not the nuclear history — the claim is about what conditions governance requires, and AI lacks the physical observability condition. Secondary claim about input-based governance (chips) is separately extractable and actionable.
## Key Facts
- Limited Test Ban Treaty (1963) works because nuclear tests produce seismically detectable explosions and atmospheric isotope signatures
- IAEA safeguards (1957+) verify declared plutonium/uranium quantities against declared facilities using isotope signatures
- Prandi et al. (bench2cop, 2025) analyzed ~195,000 benchmark questions and found zero coverage of oversight evasion, self-replication, or autonomous AI development
- EU AI Act Article 92 provides compulsory access to APIs and source code for evaluation
- Brundage et al.'s AAL framework (2026) marks AAL-3/4 (deception-resilient evaluation) as currently technically infeasible
- Nuclear governance timeline: ~23 years from Hiroshima (1945) to NPT (1968)