diff --git a/cmd/zstream/zstream_io.c b/cmd/zstream/zstream_io.c
index 596ea88cd9be..1d174a6eadf5 100644
--- a/cmd/zstream/zstream_io.c
+++ b/cmd/zstream/zstream_io.c
@@ -45,6 +45,7 @@
 #define	MEMORY_BASE_CUTOFF	(4ULL << 30)	/* 4GB */
 #define	MEMORY_PCT		10		/* % beyond the base region */
 #define	MEMORY_HYSTERESIS 	(128 << 20)	/* 128MB */
+#define	BEGIN_PAYLOAD_MAX	(1UL << 28)
 
 /*
  * Init only the filename; chain functions will prepare the FILE *
@@ -372,8 +373,13 @@ chain_read(void *item_in, void *context_in)
 	if (context->ic_offset == 0)
 		set_stream_attributes(item);
 
+	uint32_t drr_type = ATTR_IS_SET(CA_BYTESWAPPED) ?
+	    BSWAP_32(drr->drr_type) : drr->drr_type;
 	size_t payload_size = calc_payload_size(drr);
-	if (payload_size > UINT32_MAX) {
+	if (drr_type == DRR_BEGIN && payload_size > BEGIN_PAYLOAD_MAX) {
+		errx(1, "DRR_BEGIN payload too large at offset %llu",
+		    (u_longlong_t)context->ic_offset);
+	} else if (payload_size > UINT32_MAX) {
 		errx(1, "stated packet size is greater than uint32_t "
 		    "at offset %llu", (u_longlong_t)context->ic_offset);
 	} else if (payload_size > 0) {
@@ -383,8 +389,6 @@ chain_read(void *item_in, void *context_in)
 		set_payload(item, buff, payload_size);
 	}
 
-	uint32_t drr_type = ATTR_IS_SET(CA_BYTESWAPPED) ?
-	    BSWAP_32(drr->drr_type) : drr->drr_type;
 	if (drr_type >= DRR_NUMTYPES) {
 		err(1, "invalid record type %llu found at offset %llu",
 		    (u_longlong_t)drr_type, (u_longlong_t)context->ic_offset);
diff --git a/cmd/zstream/zstream_validate.c b/cmd/zstream/zstream_validate.c
index 137999578fb9..613006fc4cac 100644
--- a/cmd/zstream/zstream_validate.c
+++ b/cmd/zstream/zstream_validate.c
@@ -34,6 +34,8 @@ typedef struct {
 	int	nesting;
 	uint64_t featureflags;
 	boolean_t begin_spill;
+	boolean_t compound;
+	boolean_t concluded;
 } validate_context_t;
 
 static validate_context_t 	contexts[MAX_VALIDATIONS];
@@ -84,35 +86,55 @@ chain_validate_records(void *item_in, void *context_in)
 	if (OPTION_ENABLED(CA_DO_NOT_VALIDATE))
 		return (D_OK);
 
+	if (context->concluded) {
+		errx(1, "record after compound stream conclusion "
+		    "at offset %llu", (u_longlong_t)item->dp_stream_offset);
+	}
+
 	if (item->dp_stream_offset == 0 && drr->drr_type != DRR_BEGIN) {
 		warnx("warning - first record is not DRR_BEGIN");
 	}
 
 	if (drr->drr_type == DRR_BEGIN) {
-		VERIFY0(context->nesting);
+		if (context->nesting != 0) {
+			errx(1, "nested DRR_BEGIN record at offset %llu",
+			    (u_longlong_t)item->dp_stream_offset);
+		}
 		context->nesting++;
+		if (item->dp_stream_offset == 0) {
+			context->compound = DMU_GET_STREAM_HDRTYPE(
+			    drr->drr_u.drr_begin.drr_versioninfo) ==
+			    DMU_COMPOUNDSTREAM;
+		}
 		context->featureflags = DMU_GET_FEATUREFLAGS(
 		    drr->drr_u.drr_begin.drr_versioninfo);
 		context->begin_spill = !!(drr->drr_u.drr_begin.drr_flags &
 		    DRR_FLAG_SPILL_BLOCK);
 	} else if (drr->drr_type == DRR_END) {
 		VERIFY3S(context->nesting, >=, 0);
-		if (context->nesting > 0)
+		if (context->nesting > 0) {
 			context->nesting--;
+		} else if (context->compound &&
+		    IS_CONCLUSION(drr, DRR_END)) {
+			context->concluded = B_TRUE;
+		} else {
+			errx(1, "DRR_END record outside a stream "
+			    "at offset %llu",
+			    (u_longlong_t)item->dp_stream_offset);
+		}
 	} else if (drr->drr_type >= DRR_NUMTYPES) {
 		errx(1, "unknown record type: %d", drr->drr_type);
 	} else {
-		VERIFY3S(context->nesting, ==, 1);
+		if (context->nesting != 1) {
+			errx(1, "record outside a stream at offset %llu",
+			    (u_longlong_t)item->dp_stream_offset);
+		}
 	}
 
 	is_raw = validate_stream_has_feature(context, DMU_BACKUP_FEATURE_RAW);
 
 	switch (drr->drr_type) {
 
-	case DRR_BEGIN:
-		VERIFY3U(item->dp_payload_size, <=, 1UL << 28);
-		break;
-
 	case DRR_OBJECT:
 		err = recv_check_drr_object(drro, NULL, is_raw,
 		    context->begin_spill, context->featureflags, errbuf,
@@ -171,6 +193,8 @@ serial_validate_records(void)
 	context->nesting = 0;
 	context->featureflags = 0;
 	context->begin_spill = B_FALSE;
+	context->compound = B_FALSE;
+	context->concluded = B_FALSE;
 
 	chain_step_t step = {
 		.cs_type = CS_SERIAL,
diff --git a/tests/runfiles/common.run b/tests/runfiles/common.run
index 0f77e1bb36d9..ff085a048f01 100644
--- a/tests/runfiles/common.run
+++ b/tests/runfiles/common.run
@@ -1180,7 +1180,7 @@ tests = ['zstream_checksum_001_pos',
     'zstream_recompress_005_pos',
     'zstream_redup_001_pos',
     'zstream_selftest_queue_001_pos',
-    'zstream_validate_001_neg']
+    'zstream_validate_001_neg', 'zstream_validate_002_neg']
 tags = ['functional', 'zstream']
 
 [tests/functional/zvol/zvol_ENOSPC]
diff --git a/tests/zfs-tests/tests/Makefile.am b/tests/zfs-tests/tests/Makefile.am
index dd8123fc643d..d89ca684db7f 100644
--- a/tests/zfs-tests/tests/Makefile.am
+++ b/tests/zfs-tests/tests/Makefile.am
@@ -423,6 +423,7 @@ nobase_dist_datadir_zfs_tests_tests_DATA += \
 	functional/zstream/undersized-write-lsize.zsend.bz2 \
 	functional/zstream/big-endian-long-payloads.zsend.bz2 \
 	functional/zstream/little-endian-long-payloads.zsend.bz2 \
+	functional/zstream/make-invalid-context-streams.py \
 	functional/zstream/beadtbn-new.dump.bz2 \
 	functional/zstream/beadtbn-old.dump.bz2 \
 	functional/zstream/beadtbx-new.dump.bz2 \
@@ -2509,6 +2510,7 @@ nobase_dist_datadir_zfs_tests_tests_SCRIPTS += \
 	functional/zstream/zstream_recompress_005_pos.ksh \
 	functional/zstream/zstream_redup_001_pos.ksh \
 	functional/zstream/zstream_validate_001_neg.ksh \
+	functional/zstream/zstream_validate_002_neg.ksh \
 	functional/zstream/zstream_selftest_queue_001_pos.ksh \
 	functional/zvol/zvol_cli/cleanup.ksh \
 	functional/zvol/zvol_cli/setup.ksh \
diff --git a/tests/zfs-tests/tests/functional/zstream/make-invalid-context-streams.py b/tests/zfs-tests/tests/functional/zstream/make-invalid-context-streams.py
new file mode 100644
index 000000000000..a31b3779d2fd
--- /dev/null
+++ b/tests/zfs-tests/tests/functional/zstream/make-invalid-context-streams.py
@@ -0,0 +1,74 @@
+#!/usr/bin/env python3
+# SPDX-License-Identifier: CDDL-1.0
+#
+# This file and its contents are supplied under the terms of the
+# Common Development and Distribution License ("CDDL"), version 1.0.
+# You may only use this file in accordance with the terms of version
+# 1.0 of the CDDL.
+#
+# A full copy of the text of the CDDL should have accompanied this
+# source.  A copy of the CDDL is also available via the Internet at
+# https://opensource.org/license/CDDL-1.0.
+#
+
+import struct
+import sys
+from pathlib import Path
+
+DRR_BEGIN = 0
+DRR_OBJECT = 1
+DRR_END = 5
+DMU_BACKUP_MAGIC = 0x2F5BACBAC
+DMU_SUBSTREAM_VERSIONINFO = 0x18C0011
+DMU_COMPOUNDSTREAM_VERSIONINFO = (
+    (DMU_SUBSTREAM_VERSIONINFO & ~0x3) | 0x2
+)
+BEGIN_PAYLOAD_MAX = 1 << 28
+RECORD_SIZE = 312
+
+
+def record(byte_order, record_type, payload_size=0):
+    header = struct.pack(f"{byte_order}II", record_type, payload_size)
+    return header + bytes(RECORD_SIZE - len(header))
+
+
+def begin(byte_order, payload_size=0, compound=False):
+    toname = b"test/source@baseline".ljust(256, b"\0")
+    value = struct.pack(
+        f"{byte_order}IIQQQIIQQ256s",
+        DRR_BEGIN,
+        payload_size,
+        DMU_BACKUP_MAGIC,
+        (DMU_COMPOUNDSTREAM_VERSIONINFO if compound
+         else DMU_SUBSTREAM_VERSIONINFO),
+        0,
+        2,
+        0,
+        1,
+        0,
+        toname,
+    )
+    assert len(value) == RECORD_SIZE
+    return value
+
+
+output = Path(sys.argv[1])
+output.mkdir(parents=True, exist_ok=True)
+for byte_order, name in ((">", "big"), ("<", "little")):
+    begin_record = begin(byte_order)
+    end_record = record(byte_order, DRR_END)
+    (output / f"nested-begin-{name}.zsend").write_bytes(
+        begin_record + begin_record
+    )
+    (output / f"record-after-end-{name}.zsend").write_bytes(
+        begin_record + end_record + record(byte_order, DRR_OBJECT)
+    )
+    (output / f"extra-end-{name}.zsend").write_bytes(
+        begin_record + end_record + end_record
+    )
+    (output / f"compound-extra-end-{name}.zsend").write_bytes(
+        begin(byte_order, compound=True) + end_record + end_record + end_record
+    )
+    (output / f"oversized-begin-{name}.zsend").write_bytes(
+        begin(byte_order, BEGIN_PAYLOAD_MAX + 1)
+    )
diff --git a/tests/zfs-tests/tests/functional/zstream/zstream_validate_002_neg.ksh b/tests/zfs-tests/tests/functional/zstream/zstream_validate_002_neg.ksh
new file mode 100755
index 000000000000..f53e07ef1e60
--- /dev/null
+++ b/tests/zfs-tests/tests/functional/zstream/zstream_validate_002_neg.ksh
@@ -0,0 +1,69 @@
+#!/bin/ksh -p
+# SPDX-License-Identifier: CDDL-1.0
+
+#
+# This file and its contents are supplied under the terms of the
+# Common Development and Distribution License ("CDDL"), version 1.0.
+# You may only use this file in accordance with the terms of version
+# 1.0 of the CDDL.
+#
+# A full copy of the text of the CDDL should have accompanied this
+# source.  A copy of the CDDL is also available via the Internet at
+# https://opensource.org/license/CDDL-1.0.
+#
+
+. $STF_SUITE/tests/functional/zstream/zstream.kshlib
+
+#
+# Description:
+# Verify that zstream dump rejects invalid record context and oversized BEGIN
+# payloads while accepting valid compound streams.
+#
+# Strategy:
+# 1. Generate malformed stream records in both byte orders
+# 2. Check context and payload-size errors include the record offset
+# 3. Check existing valid compound-stream fixtures remain accepted
+#
+
+verify_runnable "both"
+
+log_assert "zstream rejects invalid record context and oversized payloads"
+
+typeset streams=$BACKDIR/context-streams
+typeset err=$BACKDIR/context-error
+typeset generator=$ZSTREAM_DATADIR/make-invalid-context-streams.py
+
+function verify_rejected # stream message
+{
+	typeset stream=$1
+	typeset message=$2
+	typeset ret
+
+	zstream dump -C "$stream" >/dev/null 2>"$err"
+	ret=$?
+	log_must test "$ret" -eq 1
+	log_must grep -q "$message" "$err"
+}
+
+log_must python3 "$generator" "$streams"
+
+for endian in big little; do
+	verify_rejected "$streams/nested-begin-$endian.zsend" \
+	    "nested DRR_BEGIN record at offset 312"
+	verify_rejected "$streams/record-after-end-$endian.zsend" \
+	    "record outside a stream at offset 624"
+	verify_rejected "$streams/extra-end-$endian.zsend" \
+	    "DRR_END record outside a stream at offset 624"
+	verify_rejected "$streams/compound-extra-end-$endian.zsend" \
+	    "record after compound stream conclusion at offset 936"
+	verify_rejected "$streams/oversized-begin-$endian.zsend" \
+	    "DRR_BEGIN payload too large at offset 0"
+done
+
+for endian in big little; do
+	typeset stream="${endian}-endian-all-drr-types-incr-XDR.zsend.bz2"
+	typeset compound="$ZSTREAM_DATADIR/$stream"
+	log_must eval "bzcat $compound | zstream dump >/dev/null"
+done
+
+log_pass "zstream reports invalid stream context with record offsets"