Embedding Forbidden Text in Spyware to Discourage AI Analysis

A set of Schneier blog posts covering multiple security and privacy topics. Key items: (1) Malware authors are embedding policy-triggering “forbidden” text (references to nuclear/biological weapons and fake system instructions) inside large JavaScript block comments at the start of payloads to poison AI-mediated analysis pipelines; the real payload follows (try{eval(...)} around a large character-code array with a ROT-style substitution), which can cause refusal, prompt confusion, context pollution, or premature benign classification in weak scanners. (2) Anthropic’s Fable 5 (a safety-hardened