More details on Fable 5’s cyber safeguards and our jailbreak framework

Claude Fable 5 is now available globally with updated cybersecurity safeguards. These safeguards use AI classifiers to monitor and block potentially dangerous requests while allowing benign activities. Because cybersecurity tools are often dual-use, the system categorizes requests into four levels: prohibited, high-risk dual use, low-risk dual use, and benign. To ensure safety, the model employs a large safety margin, which intentionally blocks some benign or low-risk requests to prevent harmful ones from slipping through.

The company is also introducing a draft framework to measure the severity of AI jailbreaks, which are methods used to bypass these safety restrictions. This Cyber Jailbreak Severity scale aims to create a consistent standard for industry and government to evaluate risks. The framework assesses jailbreaks based on four factors: the level of new capability provided to an attacker, the breadth of tasks the technique can perform, the ease of turning the jailbreak into a real-world attack, and how easily the technique can be discovered.

These measures are designed to balance the need for defensive cybersecurity research with the risk of misuse. By establishing clear definitions and a standardized severity scale, the developers hope to foster collaboration across academia, industry, and government. Security researchers are encouraged to test the model and report potential vulnerabilities through a new HackerOne program to help refine these protections over time.