Attack technique where malicious instructions are embedded in content processed by AI language models
2 stories