US Mythos and Fable ban: 'Make sure you can always switch to a different AI model'

Amsterdam, 16 Jun 2026. In this article published in Het Financieele Dagblad, Anthropic's decision to take its Mythos and Fable models offline — following a US government export control order — serves as a wake-up call for European organisations dependent on American AI. The trigger was a reported security vulnerability allowing the models' safety instructions to be bypassed, though cybersecurity expert Frank Breedijk notes that watertight protection is fundamentally impossible for large language models. Rewire CEO Wouter Huygen draws the practical lesson: organisations should build AI systems that are model-agnostic from the outset, so they can switch providers when needed.

Read the full article in Het Financieele Dagblad [in Dutch].