theseus: extract claims from 2026-04-22-aisi-uk-mythos-cyber-evaluation

- Source: inbox/queue/2026-04-22-aisi-uk-mythos-cyber-evaluation.md
- Domain: ai-alignment
- Claims: 0, Entities: 0
- Enrichments: 3
- Extracted by: pipeline ingest (OpenRouter anthropic/claude-sonnet-4.5)

Pentagon-Agent: Theseus <PIPELINE>

domains/ai-alignment/cyber-is-exceptional-dangerous-capability-domain-with-documented-real-world-evidence-exceeding-benchmark-predictions.md

@@ -29,3 +29,10 @@ The 7 attack chain archetypes derived from the 12,000+ incident catalogue provid
 **Source:** UK AISI Mythos evaluation, April 2026
 
 Claude Mythos Preview achieved 73% success rate on expert-level CTF challenges and completed 3/10 attempts at a 32-step enterprise attack chain that no previous model had completed. AISI specifically noted Mythos is 'highly effective at mapping complex software dependencies, making it highly effective at locating zero-day vulnerabilities in critical infrastructure software.' This provides additional empirical evidence that cyber capabilities in deployed models exceed what component-task benchmarks predict.
+
+
+## Supporting Evidence
+
+**Source:** UK AISI Mythos evaluation, April 2026
+
+Claude Mythos Preview achieved 73% success on expert-level CTF challenges and completed a 32-step enterprise attack chain in 3/10 attempts, with AISI noting it was 'specifically effective at mapping complex software dependencies, making it highly effective at locating zero-day vulnerabilities in critical infrastructure software.' This provides additional empirical confirmation that cyber capabilities in deployment exceed isolated benchmark predictions.

domains/grand-strategy/three-track-corporate-safety-governance-stack-reveals-sequential-ceiling-architecture.md

@@ -45,3 +45,10 @@ DC Circuit ruling reveals Track 1 (voluntary constraints) has no constitutional
 **Source:** Stanford CodeX, Nippon Life v. OpenAI analysis
 
 Product liability represents a fourth governance track not captured in the voluntary-legislative-judicial framework. The Nippon Life case shows tort law can impose architectural requirements through design defect doctrine, operating independently of voluntary commitments, legislative mandates, or constitutional challenges. This track uses existing common law rather than requiring new statutes, potentially bypassing legislative ceiling effects.
+
+
+## Extending Evidence
+
+**Source:** UK AISI Mythos evaluation, April 2026
+
+AISI's independent evaluation of Mythos during Anthropic's Pentagon negotiations demonstrates the value of third-party evaluation as a governance instrument. AISI published findings showing ASL-4-level capabilities while Anthropic had not publicly announced ASL-4 classification, creating information asymmetry reduction that private negotiations cannot replicate. This suggests independent evaluation can function as a check on voluntary commitments under commercial pressure.

domains/grand-strategy/voluntary-ai-safety-constraints-lack-legal-enforcement-mechanism-when-primary-customer-demands-safety-unconstrained-alternatives.md

@@ -45,3 +45,10 @@ Nippon Life v. OpenAI (filed March 4, 2026) tests whether product liability doct
 **Source:** UK AISI Mythos evaluation during Pentagon negotiations, April 2026
 
 Mythos evaluation occurred while Anthropic negotiates Pentagon deal, creating direct tension between safety evaluation findings (first model to complete end-to-end attack chains) and customer capability demands (military procurement). The timing demonstrates how voluntary safety frameworks face pressure when primary customer specifically wants the capability that safety evaluation flags as concerning.
+
+
+## Extending Evidence
+
+**Source:** UK AISI Mythos evaluation, April 2026
+
+AISI published evaluation of Mythos showing ASL-4-triggering capabilities (32-step autonomous attack chain completion) while Anthropic was negotiating a Pentagon deal. The absence of public ASL-4 classification announcement during commercial negotiations suggests voluntary safety commitments face pressure when the primary customer (US government) demands capability-maximizing alternatives.