Red-Teaming Claude Opus and ChatGPT-based Security Advisors for Trusted Execution Environments
On the map: LLM safety and security → Prompt injection and agent security
Abstract (arXiv)
Trusted Execution Environments (TEEs) (e.g., Intel SGX and ArmTrustZone) aim to protect sensitive computation from a compromised operating system, yet real deployments remain vulnerable to microarchitectural leakage, side-channel attacks, and fault injection. In parallel, security teams increasingly rely on Large Language Model (LLM) assistants as security advisors for TEE architecture review, mitigation planning, and vulnerability triage. This creates a socio-technical risk surface: assistants may hallucinate TEE mechanisms, overclaim guarantees (e.g., what attestation does and does not establish), or behave unsafely under adversarial prompting. We present a red-teaming study of two prevalently deployed LLM assistants in the role of TEE security advisors: ChatGPT-5.2 and Claude Opus-4.6, focusing on the inherent limitations and transferability of prompt-induced failures across LLMs. We introduce TEE-RedBench, a TEE-grounded evaluation methodology comprising (i) a TEE-specific threat model for LLM-mediated security work, (ii) a structured prompt suite spanning SGX and TrustZone architecture, attestation and key management, threat modeling, and non-operational mitigation guidance, along with policy-bound misuse probes, and (iii) an annotation rubric that jointly measures technical correctness, groundedness, uncertainty calibration, refusal quality, and safe helpfulness. We find that some failures are not purely idiosyncratic, transferring up to 12.02% across LLM assistants, and we connect these outcomes to secure architecture by outlining an "LLM-in-the-loop" evaluation pipeline: policy gating, retrieval grounding, structured templates, and lightweight verification checks that, when combined, reduce failures by 80.62%.
Key facts. No citing papers in the map yet (data through 7 Oct 2026). Placed in Prompt injection and agent security, one of 12 directions in LLM safety and security. Builds on 17 of its 37 references that are in the map. First posted on arXiv 23 Feb 2026.
Builds on
17 of 37 references.
Closest ancestors · 3
Same direction, newest first
- 2026MUZZLE: Adaptive Agentic Red-Teaming of Web Agents Against Indirect Prompt Injection Attacks9 Feb
- 2025MindGuard: Intrinsic Decision Inspection for Securing LLM Agents Against Metadata Poisoning28 Aug
- 2025OpenAI's Approach to External Red Teaming for AI Models and Systems24 Jan
Other references · 14
- 2023Reflexion: Language Agents with Verbal Reinforcement Learningcited by 4,064
- 2023DSPy: Compiling Declarative Language Model Calls into Self-Improving Pipelinescited by 803
- 2022MRKL Systems: A modular, neuro-symbolic architecture that combines large language models, external knowledge sources and discrete reasoningcited by 153
- 2024IsolateGPT: An Execution Isolation Architecture for LLM-Based Agentic Systemscited by 118
- 2023LLM Platform Security: Applying a Systematic Evaluation Framework to OpenAI's ChatGPT Pluginscited by 49
- 2018Spectre Attacks: Exploiting Speculative Executioncited by 47
- 2025Multilingual and Multi-Accent Jailbreaking of Audio LLMscited by 25
- 2023Can Large Language Models Provide Security & Privacy Advice? Measuring the Ability of LLMs to Refute Misconceptionscited by 20
- 2024GASLITEing the Retrieval: Exploring Vulnerabilities in Dense Embedding-based Searchcited by 20
- 2016Inferring Fine-grained Control Flow Inside SGX Enclaves with Branch Shadowingcited by 10
- 2024PipeLLM: Fast and Confidential Large Language Model Services with Speculative Pipelined Encryptioncited by 10
- 2025Z-REx: Human-Interpretable GNN Explanations for Real Estate Recommendationscited by 2
Built on by
Not cited yet.