Anthropic has pushed back against claims that its new Claude Fable 5 AI model was jailbroken by a researcher known as Pliny the Liberator. The company says the researcher's demonstration doesn't represent a true bypass of its safety systems, which include independent classifiers that prevent the model from helping with dangerous tasks like bioweapons development or sophisticated cyberattacks. Anthropic maintains that the examples shown only elicited general public information and didn't circumvent the strongest protections built into the system.