Constitutional Classifiers++: Efficient Production-Grade Defenses against Universal Jailbreaks

We introduce enhanced Constitutional Classifiers that deliver production-grade jailbreak robustness with dramatically reduced computational costs and refusal rates compared to previous-generation defenses. Our system combines several key insights. First, we develop exchange classifiers that evaluate model responses in their full conversational context, which addresses vulnerabilities in last-generation systems that examine outputs in isolation. Second, we implement a two-stage classifier cascade where lightweight classifiers screen all traffic and escalate only suspicious exchanges to more expensive classifiers. Third, we train efficient linear probe classifiers and ensemble them with external classifiers to simultaneously improve robustness and reduce computational costs. Together, these techniques yield a production-grade system achieving a 40x computational cost reduction compared to our baseline exchange classifier, while maintaining a 0.05% refusal rate on production traffic. Through extensive red-teaming comprising over 1,700 hours, we demonstrate strong protection against universal jailbreaks -- no attack on this system successfully elicited responses to all eight target queries comparable in detail to an undefended model. Our work establishes Constitutional Classifiers as practical and efficient safeguards for large language models.

Key Contributions

• Exchange classifiers that evaluate model responses in full conversational context, addressing reconstruction and obfuscation attacks that defeated output-only classifiers
• Two-stage classifier cascade using lightweight linear probe classifiers as a first stage to reduce compute overhead by 40x while maintaining 0.05% refusal rate
• Linear activation probe classifiers with logit smoothing and weighted softmax loss, which can be ensembled with external classifiers for improved robustness at negligible cost

🛡️ Threat Analysis

Details

Similar Papers