diff --git a/lib/rbs/ast/ruby/comment_block.rb b/lib/rbs/ast/ruby/comment_block.rb
index 1eaaa4a399..14de35ecb7 100644
--- a/lib/rbs/ast/ruby/comment_block.rb
+++ b/lib/rbs/ast/ruby/comment_block.rb
@@ -199,7 +199,7 @@ def parse_annotation_lines(start_line, end_line, variables)
         def trailing_annotation(variables)
           if trailing?
             comment = comments[0] or raise
-            if comment.location.slice.start_with?(/#[:\[]/)
+            if comment.location.slice.start_with?("#:", "#[")
               begin
                 Parser.parse_inline_trailing_annotation(comment_buffer, 0...comment_buffer.last_position, variables: variables)
               rescue ParsingError => error
diff --git a/test/rbs/inline_parser_test.rb b/test/rbs/inline_parser_test.rb
index befd885f32..a6de530c46 100644
--- a/test/rbs/inline_parser_test.rb
+++ b/test/rbs/inline_parser_test.rb
@@ -1459,6 +1459,18 @@ class Foo
     end
   end
 
+  def test_error__attr_type_with_invalid_utf8
+    omit_on_truffle_ruby! "The C extension does not raise `RBS::ParsingError` for invalid UTF-8"
+    omit_on_jruby! "`Prism::Location#start_line_slice` returns `nil` on JRuby's prism"
+
+    ["\xDD", "\xFF", "\xC3", "\xE2\x82"].each do |invalid_bytes|
+      source = "class Foo\n  attr_reader :value #: Integer # #{invalid_bytes}\nend\n".dup.force_encoding(Encoding::UTF_8)
+      result = parse(source)
+
+      assert_instance_of RBS::InlineParser::Diagnostic::AnnotationSyntaxError, result.diagnostics.fetch(0)
+    end
+  end
+
   def test_parse__class_with_super_class
     result = parse(<<~RUBY)
       class Child < Parent