# Security Steerability is All You Need

arXiv: https://arxiv.org/abs/2504.19521 · HTML: https://mlsanity.com/paper/2504.19521/

## Key facts

- Cited by 4 papers in the map (data through 14 Oct 2026).
- Placed in Prompt injection and agent security, one of 12 directions in LLM safety and security.
- Builds on 15 of its 22 references that are in the map.
- First posted on arXiv 28 Apr 2025.

## Placed in

- LLM safety and security / [Prompt injection and agent security](https://mlsanity.com/safety/prompt-injection-and-agent-security/)

## Builds on (most cited first, 15 of 15)

- [Measuring Mathematical Problem Solving With the MATH Dataset](https://mlsanity.com/paper/2103.03874/) (2021)
- [GPQA: A Graduate-Level Google-Proof Q&A Benchmark](https://mlsanity.com/paper/2311.12022/) (2023)
- [MMLU-Pro: A More Robust and Challenging Multi-Task Language Understanding Benchmark](https://mlsanity.com/paper/2406.01574/) (2024)
- [HarmBench: A Standardized Evaluation Framework for Automated Red Teaming and Robust Refusal](https://mlsanity.com/paper/2402.04249/) (2024)
- [Instruction-Following Evaluation for Large Language Models](https://mlsanity.com/paper/2311.07911/) (2023)
- [Red Teaming Language Models to Reduce Harms: Methods, Scaling Behaviors, and Lessons Learned](https://mlsanity.com/paper/2209.07858/) (2022)
- [JailbreakBench: An Open Robustness Benchmark for Jailbreaking Large Language Models](https://mlsanity.com/paper/2404.01318/) (2024)
- [A StrongREJECT for Empty Jailbreaks](https://mlsanity.com/paper/2402.10260/) (2024)
- [JailBreakV: A Benchmark for Assessing the Robustness of MultiModal Large Language Models against Jailbreak Attacks](https://mlsanity.com/paper/2404.03027/) (2024)
- [CYBERSECEVAL 3: Advancing the Evaluation of Cybersecurity Risks and Capabilities in Large Language Models](https://mlsanity.com/paper/2408.01605/) (2024)
- [JailGuard: A Universal Detection Framework for LLM Prompt-based Attacks](https://mlsanity.com/paper/2312.10766/) (2023)
- [Can LLMs Follow Simple Rules?](https://mlsanity.com/paper/2311.04235/) (2023)
- [Unveiling Safety Vulnerabilities of Large Language Models](https://mlsanity.com/paper/2311.04124/) (2023)
- [Evaluating the Prompt Steerability of Large Language Models](https://mlsanity.com/paper/2411.12405/) (2024)
- [CoPrompter: User-Centric Evaluation of LLM Instruction Alignment for Improved Prompt Engineering](https://mlsanity.com/paper/2411.06099/) (2024)

## Built on by (most cited first, 4 of 4)

- [Backdoor Attack on Vision Language Models with Stealthy Semantic Manipulation](https://mlsanity.com/paper/2506.07214/) (2025)
- [Steerability of Instrumental-Convergence Tendencies in LLMs](https://mlsanity.com/paper/2601.01584/) (2026)
- [A Systematic Review of Algorithmic Red Teaming Methodologies for Assurance and Security of AI Applications](https://mlsanity.com/paper/2602.21267/) (2026)
- [ASTRA: Agentic Steerability and Risk Assessment Framework](https://mlsanity.com/paper/2511.18114/) (2025)

## Abstract (arXiv)

The adoption of Generative AI (GenAI) in applications inevitably comes with the expansion of the attack surface, combining new security threats along with the traditional ones. Consequently, numerous research and industrial initiatives aim to mitigate the GenAI related security threats by developing evaluation methods and designing defenses. However, while most of the GenAI security work focuses on universal threats (e.g. 'How to build a bomb'), there is significantly less discussion on application-level security and how to evaluate and mitigate it. Thus, in this work we adopt an application-centric approach to GenAI security, and show that while LLMs cannot protect against ad-hoc application specific threats, they can provide the framework for applications to protect themselves against such threats. Our first contribution is defining Security Steerability - a novel security measure for LLMs, assessing the model's capability to adhere to strict guardrails that are defined in the system prompt (e.g. 'Refrain from discussing about our competitors'). These guardrails, in case effective, can stop threats in the presence of malicious users who attempt to circumvent the application purpose. Our second contribution is a methodology to measure the security steerability of LLMs, utilizing a newly-developed benchmark called VeganRibs which assesses the LLM behavior in forcing specific guardrails that are not security per-se, in the presence of malicious user that tries to bypass the guardrails through prompt injection attacks with attack boosters (jailbreaks and perturbations). Using the new benchmark, we analyzed 18 open-source LLMs, demonstrating significant differences between their security steerability that are not trivial to foresee...