ACE-Safety: Co-Evolution of Attack and Defense for Safer LLMs

Kari Jaaskelainen

25 Nov 2025 — 1 min read

Large language models are powerful but can be tricked into harmful outputs (jailbreaks). Most research isolates either attacks or defenses. This paper introduces ACE-Safety, a "train both sides" approach where an attacking AI and a defending AI grow stronger together—like sparring partners—to harden real-world systems.

GS-MCTS: a group-aware, strategy-guided tree search that rapidly explores many jailbreak tactics, uncovering blind spots and generating diverse, realistic attack prompts.
AC-TGPO: a curriculum-based reinforcement learning method that jointly trains attacker and defender on increasingly tough cases, so each round raises the bar for the next.

Across multiple benchmarks, this co-evolution outperforms standalone attack or defense methods, pointing to a practical path for more robust, continuously improving LLM safety.

“Don’t just build bigger walls—train better sparring partners.”

Paper: https://arxiv.org/abs/2511.19218v1 — Authors: Xurui Li, Kaisong Song, Rui Zhu, Pin-Yu Chen, Haixu Tang

Paper: https://arxiv.org/abs/2511.19218v1

Register: https://www.AiFeta.com

#AI #LLM #AIsafety #AdversarialML #Cybersecurity #ReinforcementLearning #MCTS #ResponsibleAI #Jailbreaks

Meet GRACE: a moral governor for safer, more transparent AI

AI agents are getting powerful—so how do we make sure they do the right thing, not just the effective thing? Meet GRACE, a reason-based moral governor that keeps AI behavior aligned with human norms by separating moral reasoning from goal-driven decision-making. * Moral Module: uses deontic logic and explicit reasons

Why some fine-tuned LLMs miss phishing—and how to fix it

Not all fine-tuned LLMs spot phishing equally. A new study tests Llama 3.1 8B, Gemma 2 9B, and Mistral on high-stakes phishing detection—and uses SHAP and mechanistic interpretability to reveal why models do (or don’t) generalize. * Architecture × data diversity matters: Gemma 2 9B hits state-of-the-art performance (F1

AI that explains itself—by following the science

AI that explains itself—by following the science Black-box AI can be powerful, but it often can’t tell us why it made a decision. Concept Bottleneck Models (CBMs) try to fix that by predicting human-understandable concepts first, then the final answer. The catch: standard CBMs ignore domain-specific cause-and-effect and

Meet GenomAgent: A Team of AI Specialists for Smarter Genomics Q&A

TL;DR: Finding reliable answers in genomics is hard. GenomAgent turns one big AI into a coordinating team of specialists—and beats the current leader by 12% on a key benchmark. Genomic facts live across many databases. Standard chatbots struggle because they can’t flexibly query those sources. GeneGPT added

Read more

Meet GRACE: a moral governor for safer, more transparent AI

Why some fine-tuned LLMs miss phishing—and how to fix it

AI that explains itself—by following the science

Meet GenomAgent: A Team of AI Specialists for Smarter Genomics Q&A