mlsanityData to 7 Oct 2026

DynaGuard: A Dynamic Guardian Model With User-Defined Policies

arXiv 2509.02563 ↗First posted 2 Sep 2025cs.LG cs.CLcited by 16

On the map: LLM safety and security → Safety alignment, refusal and guardrails

Abstract (arXiv)

Guardian models play a crucial role in ensuring the safety and ethical behavior of user-facing AI applications by enforcing guardrails and detecting harmful content. While standard guardian models are limited to predefined, static harm categories, we introduce DynaGuard, a suite of dynamic guardian models offering novel flexibility by evaluating text based on user-defined policies, and DynaBench, a dataset for training and evaluating dynamic guardian models. Our models provide both rapid detection of policy violations and a chain-of-thought reasoning option that articulate and justify model outputs. Critically, DynaGuard not only surpasses static models in detection accuracy on traditional safety categories, but is competitive with frontier reasoning models on free-form policy violations, all in a fraction of the time. This makes DynaGuard an critical tool for language model guardrails.

Key facts. Cited by 16 papers in the map (data through 7 Oct 2026). Placed in Safety alignment, refusal and guardrails, one of 12 directions in LLM safety and security. Builds on 20 of its 29 references that are in the map. First posted on arXiv 2 Sep 2025.

Builds on

20 of 29 references.

Closest ancestors · 7

Same direction, newest first

  1. 2025OMNIGUARD: An Efficient Approach for AI Safety Moderation Across Languages and Modalities29 May
  2. 2025GuardReasoner: Towards Reasoning-based LLM Safeguards30 Jan
  3. 2025Refining Input Guardrails: Enhancing LLM-as-a-Judge Efficiency Through Chain-of-Thought Fine-Tuning and Alignment22 Jan
  4. 2025Aegis2.0: A Diverse AI Safety Dataset and Risks Taxonomy for Alignment of LLM Guardrails15 Jan
  5. 2024Controllable Safety Alignment: Inference-Time Adaptation to Diverse Safety Requirements11 Oct
  6. 2024WildGuard: Open One-Stop Moderation Tools for Safety Risks, Jailbreaks, and Refusals of LLMs26 Jun
  7. 2024Building Guardrails for Large Language Models2 Feb

Other references · 13

  1. 2022Chain-of-Thought Prompting Elicits Reasoning in Large Language Modelscited by 12,302
  2. 2024GPT-4o System Cardcited by 4,467
  3. 2024HybridFlow: A Flexible and Efficient RLHF Frameworkcited by 1,980
  4. 2024HarmBench: A Standardized Evaluation Framework for Automated Red Teaming and Robust Refusalcited by 1,330
  5. 2023Llama Guard: LLM-based Input-Output Safeguard for Human-AI Conversationscited by 1,124
  6. 2023BeaverTails: Towards Improved Safety Alignment of LLM via a Human-Preference Datasetcited by 828
  7. 2023XSTest: A Test Suite for Identifying Exaggerated Safety Behaviours in Large Language Modelscited by 477
  8. 2023ToxicChat: Unveiling Hidden Challenges of Toxicity Detection in Real-World User-AI Conversationcited by 224
  9. 2024ShieldGemma: Generative AI Content Moderation Based on Gemmacited by 224
  10. 2024Llama Guard 3 Vision: Safeguarding Human-AI Image Understanding Conversationscited by 118
  11. 2024GuardAgent: Safeguard LLM Agents by a Guard Agent via Knowledge-Enabled Reasoningcited by 113
  12. 2024RigorLLM: Resilient Guardrails for Large Language Models against Undesired Contentcited by 65
1 more
  1. 2024CantTalkAboutThis: Aligning Language Models to Stay on Topic in Dialoguescited by 15

Built on by

Cited by year (2026 to 7 Oct)

Citing papers per year: 2025: 2; 2026: 14.

16 papers.

Sort:

By area:

  1. 2025Training-Free Policy Violation Detection via Activation-Space Whitening in LLMscited by 6
  2. 2026UbuntuGuard: A Culturally-Grounded Policy Benchmark for Equitable AI Safety in African Languagescited by 5
  3. 2026LPG: Balancing Efficiency and Policy Reasoning in Latent Policy Guardrailscited by 4
  4. 2025ProGuard: Towards Proactive Multimodal Safeguardcited by 2
  5. 2026RvB: Automating AI System Hardening via Iterative Red-Blue Gamescited by 2
  6. 2026Predict, Don't React: Value-Based Safety Forecasting for LLM Streamingcited by 2
  7. 2026SafePyramid: A Hierarchical Benchmark for In-context Policy Guardrailingcited by 1
  8. 2026What Do Compliance Detectors Read? An Audit of Activation Probes and Guard Modelscited by 1
  9. 2026BARRED: Synthetic Training of Custom Policy Guardrails via Asymmetric Debatecited by 0
  10. 2026Configurable Reward Model for Balanced Safety Alignmentcited by 0
  11. 2026PolicyShiftGuard: Benchmarking and Improving Policy-Adaptive Image Guardrailscited by 0
  12. 2026Nemotron 3.5 Content Safety Moderator: A Compact Multimodal, Multilingual, and Reasoning Enabled Content Safety Moderatorcited by 0
4 more
  1. 2026FIRE: Failure-Informed Runtime Engineering for Reliable Language-Model Agentscited by 0
  2. 2026AdaGuard: An Adaptive Guard Model with User-defined Policiescited by 0
  3. 2026CoDeL: Co-Evolutionary Defense against Indirect Prompt Injection in LLM-based Agentscited by 0
  4. 2026SafeCoEvo: Co-Evolving Safety Harnesses and Guards for LLM Agents at Test-Timecited by 0

Similar, not cited

Close in wording, no citation either way

  1. 2026CourtGuard: A Model-Agnostic Framework for Zero-Shot Policy Adaptation in LLM Safety
  2. 2025AprielGuard
  3. 2025RSafe: Incentivizing proactive reasoning to build robust and adaptive LLM safeguards
  4. 2026PL-Guard: Probabilistic Logic Reasoning for LLM Guardrails
  5. 2026TraceSafe: A Systematic Assessment of LLM Guardrails on Multi-Step Tool-Calling Trajectories
  6. 2025Safety Through Reasoning: An Empirical Study of Reasoning Guardrail Models
  7. 2025ThinkGuard: Deliberative Slow Thinking Leads to Cautious Guardrails
  8. 2025OmniGuard: Unified Omni-Modal Guardrails with Deliberate Reasoning