Thilo Hagendorff, Erik Derner, Nuria Oliver. "Large reasoning models are autonomous jailbreak agents." Nature Communications (2026). https://doi.org/10.1038/s41467-026-69010-1