Strengthening our safeguards with frontier automated red-teaming
Nate Rahn*, Chawin Sitawarin*, David Huang*, Richard Zhuang, Yousef Nassar, Sunny Duan, Erik Jones, Milad Nasr, Nicholas Carlini, Andreas Terzis, Javier Rando, 2026
We built Anvil, an automated red-teaming system that helps us strengthen our safeguards before we deploy them.
Scaling Misuse Detection with Self-Assessment
Yiming Zhang, Chen Xia, Andy Dau, John Wang, James Yu, Bobby Chen, Andreas Terzis, Nicholas Carlini, Erik Jones, 2026
Self-assessment lets Claude assess the safety of its own responses. It scales well with model capabilities and helps us identify emergent harms.