Anthropic released its unreleased Mythos model to Apple and Amazon. The partnership aims to test the model's resilience against cyberattacks before wider rollout. Developers can use the model to harden security pipelines. Anthropic’s move follows industry concerns over AI-driven vulnerabilities and signals a shift toward proactive security testing. The test will include simulated phishing and data exfiltration scenarios.