diff --git a/lib/Regex/Executor.cpp b/lib/Regex/Executor.cpp
index a8fea998b58..e2d7d50c49d 100644
--- a/lib/Regex/Executor.cpp
+++ b/lib/Regex/Executor.cpp
@@ -1048,7 +1048,6 @@ auto Context<Traits>::match(State<Traits> *s, bool onlyAtStart)
 
         case Opcode::U16MatchCharICase32: {
           const auto *insn = llvh::cast<U16MatchCharICase32Insn>(base);
-          assert(insn->c >= 0x010000 && "Character should be astral");
           bool matched = false;
           if (!c.atEnd()) {
             CodePoint cp = c.consumeUTF16();
diff --git a/test/hermes/regexp_unicode.js b/test/hermes/regexp_unicode.js
index 853f1d72952..bc817383ba3 100644
--- a/test/hermes/regexp_unicode.js
+++ b/test/hermes/regexp_unicode.js
@@ -89,6 +89,13 @@ print(/.*/u.exec("\u0101bc\ndef")[0].length);
 // We should not match a low surrogate in a Unicode regexp.
 print(!! /\uDE42/u.exec("\uD83D\uDE42ZZZ"));
 // CHECK-NEXT: false
+// We should match an unpaired surrogate.
+print(!! /\uDC00/u.exec("\uDC00"));
+// CHECK-NEXT: true
+// Test the case insensitive variant.
+print(!! /\uDC00/iu.exec("\uDC00"));
+// CHECK-NEXT: true
+// We should match the low surrogate when Unicode is off.
 print(!! /\uDE42/.exec("\uD83D\uDE42ZZZ"));
 // CHECK-NEXT: true