Amazon testing Anthropic models but reported jailbreak failures, triggering export controls
As Anthropic's trusted partner, Amazon tested Mythos/Fable with guardrails, found failures, and reported to Commerce Department, demonstrating responsible AI safety practices but…