U+00A0 | NO-BREAK SPACE (UTF-8: C2 A0) | Emitted by HTML entities, macOS Option+Space, Google Docs, and LLM markdown formatters. | Looks identical to ASCII U+0020 (20) but breaks YAML indentation, CLI flags, SQL queries, and Python/JS lexers. |
U+202F | NARROW NO-BREAK SPACE (UTF-8: E2 80 AF) | Placed before punctuation, units, and AM/PM timestamps by ChatGPT, French/ICU locale formatters, and macOS. | Fails strict string equality checks against standard spaces in unit tests and CSV importers. |
U+200B | ZERO WIDTH SPACE (UTF-8: E2 80 8B) | Left behind by contenteditable web boxes, table cell copies, and chat message copy buttons. | Creates phantom length mismatches where string.length exceeds visible character count. |
U+202E | RIGHT-TO-LEFT OVERRIDE (UTF-8: E2 80 AE) | Directional formatting control abused in Trojan-Source code attacks to reverse rendered character order. | Makes source code or filenames display differently to human reviewers than to compilers and interpreters. |
U+E0020–U+E007E | PLANE-14 TAG BYTES (UTF-8: F3 A0 80 A0+) | Four-byte UTF-8 sequences used for regional flag subtags or hidden prompt-injection / watermark strings. | Passes unseen through visual code reviews while adding 4 UTF-8 bytes (2 UTF-16 code units) per character. |