Hidden instructions that trick an AI into doing things it should not.
Korpalis selects material that helps builders and business teams understand what changed, why it matters and what to check next. Every entry below includes an original explanation and a direct link to its publisher.
A researcher successfully broke Claude Code's auto mode defenses, which Anthropic designed to block prompt injection attacks in agentic workflows. The demonstration shows that safety mechanisms intended to protect autonomous agents remain vulnerable to determined adversaries.
Researchers demonstrate a transferable adversarial attack against code language models used in search tools by perturbing variable and function names while preserving code functionality, revealing how seemingly non-functional textual changes can manipulate neural retrieval systems. The attack generalizes across programming languages and between different models, highlighting a semantic-level vulnerability in code understanding.
Invisible HTML markup can be injected into emails to deceive AI summarization tools into generating false or malicious summaries, exploiting the models' inability to detect content hidden from human readers.
The study identifies injection vulnerabilities in LLM agents where untrusted external data can be dynamically interpreted as behavior-guiding instructions, allowing attackers to subvert agent decisions. The work proposes methods to localize and defend against these dynamic injections that occur during inference rather than at static input/output boundaries.
The paper compares natural-language safety rules written in CLAUDE.md files against Claude Code's built-in enforcement mechanisms, finding a significant gap between instructional constraints and executable controls. By analyzing 481 public files and validating findings through security practitioner review, the authors quantify how many stated security goals lack corresponding technical enforcement.
A research paper proposes using blockchain technology to verify the trustworthiness of documents retrieved in RAG systems by implementing a committee-based credibility scoring mechanism. This addresses the challenge of ensuring retrieved information has not been tampered with and comes from reliable sources, especially critical for high-stakes applications.