diff --git a/Lib/test/test_tokenize.py b/Lib/test/test_tokenize.py
index ab53a20cff5539..8c91e271f4d5e8 100644
--- a/Lib/test/test_tokenize.py
+++ b/Lib/test/test_tokenize.py
@@ -3188,6 +3188,16 @@ def get_tokens(string):
             with self.subTest(case=case):
                 self.assertRaises(tokenize.TokenError, get_tokens, case)
 
+    def test_tstring_multiline_bang_underflow(self):
+        # gh-149183: t-string with '!' across two lines used to raise
+        # MemoryError because last_expr_end > last_expr_size produced a
+        # negative length that was cast to a huge size_t.
+        self.assertRaises(
+            tokenize.TokenError,
+            list,
+            tokenize.tokenize(BytesIO(b't"{!\n!x').readline),
+        )
+
     @support.skip_wasi_stack_overflow()
     def test_max_indent(self):
         MAXINDENT = 100
diff --git a/Parser/lexer/lexer.c b/Parser/lexer/lexer.c
index 7f25afec302c22..748df54338f6a3 100644
--- a/Parser/lexer/lexer.c
+++ b/Parser/lexer/lexer.c
@@ -121,12 +121,22 @@ set_ftstring_expr(struct tok_state* tok, struct token *token, char c) {
     }
     PyObject *res = NULL;
 
+    Py_ssize_t expr_len = tok_mode->last_expr_size - tok_mode->last_expr_end;
+    if (expr_len < 0) {
+        /* last_expr_end > last_expr_size: happens when '{' and the closing
+           delimiter span different source lines, causing the strlen-based
+           size tracking to underflow.  Treat as a tokenizer error rather
+           than passing a negative length (cast to huge size_t) to malloc or
+           PyUnicode_DecodeUTF8. */
+        return -1;
+    }
+
     // Look for a # character outside of string literals
     int hash_detected = 0;
     int in_string = 0;
     char quote_char = 0;
 
-    for (Py_ssize_t i = 0; i < tok_mode->last_expr_size - tok_mode->last_expr_end; i++) {
+    for (Py_ssize_t i = 0; i < expr_len; i++) {
         char ch = tok_mode->last_expr_buffer[i];
 
         // Skip escaped characters
@@ -163,7 +173,7 @@ set_ftstring_expr(struct tok_state* tok, struct token *token, char c) {
     // If we found a # character in the expression, we need to handle comments
     if (hash_detected) {
         // Allocate buffer for processed result
-        char *result = (char *)PyMem_Malloc((tok_mode->last_expr_size - tok_mode->last_expr_end + 1) * sizeof(char));
+        char *result = (char *)PyMem_Malloc((expr_len + 1) * sizeof(char));
         if (!result) {
             return -1;
         }
@@ -174,7 +184,7 @@ set_ftstring_expr(struct tok_state* tok, struct token *token, char c) {
         quote_char = 0;    // Current string quote char
 
         // Process each character
-        while (i < tok_mode->last_expr_size - tok_mode->last_expr_end) {
+        while (i < expr_len) {
             char ch = tok_mode->last_expr_buffer[i];
 
             // Handle string quotes
@@ -190,11 +200,10 @@ set_ftstring_expr(struct tok_state* tok, struct token *token, char c) {
             }
             // Skip comments
             else if (ch == '#' && !in_string) {
-                while (i < tok_mode->last_expr_size - tok_mode->last_expr_end &&
-                       tok_mode->last_expr_buffer[i] != '\n') {
+                while (i < expr_len && tok_mode->last_expr_buffer[i] != '\n') {
                     i++;
                 }
-                if (i < tok_mode->last_expr_size - tok_mode->last_expr_end) {
+                if (i < expr_len) {
                     result[j++] = '\n';
                 }
             }
@@ -211,7 +220,7 @@ set_ftstring_expr(struct tok_state* tok, struct token *token, char c) {
     } else {
         res = PyUnicode_DecodeUTF8(
             tok_mode->last_expr_buffer,
-            tok_mode->last_expr_size - tok_mode->last_expr_end,
+            expr_len,
             NULL
         );
     }