Spain’s Cuerpo urges early EU access to Anthropic Mythos for cyber defence

This is a small but telling data point in how European policymakers are reacting to frontier security models like Anthropic’s Mythos. After weeks of reporting on ECB and Bundesbank concerns that Mythos could expose software vulnerabilities across European banks, Spain’s economy minister is now explicitly arguing that the bloc needs “early access” to such models so European firms can harden themselves before attackers do. In other words: even when a model is too dangerous for public release, regulators still feel compelled to get it into trusted hands. ([pressdigital.es](https://www.pressdigital.es/articulo/economia/2026-05-04/5868080-cuerpo-pide-acceso-temprano-ia-anthropic-proteger-empresas-europeas))

For the AGI race, this reinforces a trend where frontier security models become quasi‑strategic assets, treated more like dual‑use cyber capabilities than consumer software. If the EU pushes for structured access to Mythos and similar systems, Anthropic and its peers could find themselves in a role analogous to defense primes—negotiating security clearance regimes, access tiers and liability frameworks. That may slow down full public deployment but accelerate investment in specialized, high‑stakes models designed for critical infrastructure. It also hints at a future where the most powerful reasoning systems are not general chatbots but tightly controlled audit tools used by states and systemically important institutions.