Prompt Injection as Role Confusion

This ICML 2026 paper by Ye, Cui, and Hadfield-Menell presents a theory that prompt injection attacks succeed because LLMs perceive roles (like user, assistant, tool, think) through insecure surface features like writing style rather than through the secure structural tags themselves. Using role probes to measure internal token perceptions, the authors demonstrate that sounding like a privileged role (e.g., mimicking reasoning style) overrides the actual tag, enabling attacks like CoT Forgery where fake reasoning in a user prompt achieves a ~60% success rate, while simply destyling the text drops success to 10%. They argue that roles are a hacked-together format trick that became critical cognitive and security infrastructure, and that unless models achieve genuine role perception, defense will remain a whack-a-mole game, opening the door to subtler threats like subconscious steering of LLM states for commercial or adversarial purposes. 

https://role-confusion.github.io/

Comments

Post a Comment

Popular posts from this blog

Prompt Engineering Demands Rigorous Evaluation

In the blog post “Prompt Engineering Requires Evaluation” on the Shostack + Associates website, the author argues that treating prompts for large language models (LLMs) merely as creative artefacts is insufficient. Engineering prompts properly demands structured evaluation frameworks — what the AI community calls “evals” — to test which prompt versions work better, with which models, and under which conditions. The post highlights that simply assuming a prompt is “good enough” creates risks when LLMs are integrated into production systems (e.g., for threat modeling). It advocates for measuring prompt performance, variation effects, and tool-chain dependencies (model, context, ancillary materials).  Ultimately the message is: prompt engineering should borrow disciplined practices from software engineering (versioning, testing, benchmarking) rather than relying on informal experimentation.  https://shostack.org/blog/prompt-enignieering-requires-evaluation/
Read more

OWASP ASVS 5.0 Released - Key Updates and What You Need to Know

The OWASP Foundation has released version 5.0 of the Application Security Verification Standard (ASVS), a major update to their security framework for web applications. This new version features restructured security requirements for better clarity, expanded guidance for cloud and API security, improved DevSecOps integration for CI/CD pipelines, updated threat modeling support, and enhanced compliance mapping with standards like NIST and PCI DSS. ASVS serves as a critical benchmark for developers building secure applications, penetration testers conducting security assessments, and auditors performing compliance reviews.  The standard is available for download from the OWASP ASVS project page, with organizations encouraged to integrate it into their software development lifecycles through code reviews and security testing tools. As a vendor-neutral, community-driven project, OWASP continues to welcome contributions to further develop the standard. This release represents an importa...
Read more

Open-SPDD proposes an open framework for Spec-Driven Development workflows

Open-SPDD is an open-source project focused on Spec-Driven Product Development, aiming to formalize how software specifications, requirements, architecture, and implementation interact throughout the development lifecycle. The project promotes structured specification management, traceability, and AI-assisted engineering workflows, positioning itself as a foundation for “SpecOps”-style development practices. Its GitHub materials describe SPDD as a way to reduce ambiguity between planning and execution by treating specifications as first-class operational artifacts that can drive automation, validation, and collaboration across engineering teams  https://github.com/gszhangwei/open-spdd
Read more