June 9, 2026: Anthropic launches Claude Fable 5 — the most capable AI model ever made available for general use, state-of-the-art on nearly all benchmarks, with major advances in software engineering, knowledge work, and scientific research. Three days later, on June 12 at 5:21 PM ET, Anthropic receives an export control directive from the US government: all access for foreign nationals — including Anthropic's own foreign national employees — must be suspended immediately. Fable 5 and Mythos 5 are abruptly disabled for all customers worldwide. Anthropic complied but publicly objected: the demonstrated jailbreak technique involved known, minor vulnerabilities also discoverable via other publicly available models including OpenAI's GPT-5.5. This is the first time an AI model itself has been treated as strategic export-controlled technology.
Michael and Nadine work through four concrete scenarios that emerge the moment the block hits: an EU AI Act audit where a regulator asks which AI version supported which decision; legal proceedings where a customer disputes an AI-driven outcome; vendor dependency and executive liability under NIS2, DORA, and the Dutch Cybersecurity Act; and a cyber insurance claim requiring proof of operational impact. In every scenario the core question is the same: can you prove what actually happened — even when the model is no longer available?
Attesto does not prevent a model from disappearing or a government from issuing export controls. But Attesto ensures that every AI interaction — model identity, version, timestamp, prompt hash, output hash, digital signature — is cryptographically recorded beyond dispute via the Proof of Evolution system. The cryptographic black box above AI systems. The Nova/IVC layer is already running live in production.