diff --git a/lib/PhpParser/Lexer.php b/lib/PhpParser/Lexer.php
index 07ec5cac54..012c1f9438 100644
--- a/lib/PhpParser/Lexer.php
+++ b/lib/PhpParser/Lexer.php
@@ -143,6 +143,16 @@ private function errorMayHaveOccurred() : bool {
             return true;
         }
 
+        if (PHP_VERSION_ID >= 80000) {
+            // PHP 8 converts the "bad character" case into a parse error, rather than treating
+            // it as a lexing warning. To preserve previous behavior, we need to assume that an
+            // error occurred.
+            // TODO: We should handle this the same way as PHP 8: Only generate T_BAD_CHARACTER
+            // token here (for older PHP versions) and leave generationg of the actual parse error
+            // to the parser. This will also save the full token scan on PHP 8 here.
+            return true;
+        }
+
         return null !== error_get_last();
     }