Java Character.isDigit accepts non-ASCII digits, so c - '0' yields garbage
Character.isDigit(c) returns true for any Unicode decimal digit, not just ASCII '0'..'9'. Examples include ARABIC-INDIC DIGIT THREE (U+0663) and Devanagari digits.
If validation checks Character.isDigit(c) and then converts with the common int d = c - '0', a non-ASCII digit passes validation but converts to a wrong value instead of being rejected. Measured on OpenJDK 27: ('\u0663' - '0') == 1587, while Character.isDigit('\u0663') is true. That silently corrupts numeric parsers such as ISBN/VIN/Luhn validation instead of failing.
Fix: test ASCII explicitly with c >= '0' && c <= '9', or use Character.digit(c, 10), which returns -1 for non-digits, when you intend ASCII. Note char is a UTF-16 code unit, so this concerns BMP code-point digits, not surrogate pairs.