diff --git a/sqlite_utils/cli.py b/sqlite_utils/cli.py
index b7de4191..5c772c88 100644
--- a/sqlite_utils/cli.py
+++ b/sqlite_utils/cli.py
@@ -1,7 +1,7 @@
 import base64
 import click
 from click_default_group import DefaultGroup  # type: ignore
-from datetime import datetime
+from datetime import datetime, timezone
 import hashlib
 import pathlib
 from runpy import run_module
@@ -865,7 +865,7 @@ def inner(fn):
                     required=True,
                 ),
                 click.argument("table"),
-                click.argument("file", type=click.File("rb"), required=True),
+                click.argument("file", type=click.File("rb", lazy=True), required=True),
                 click.option(
                     "--pk",
                     help="Columns to use as the primary key, e.g. id",
@@ -1949,12 +1949,16 @@ def memory(
             fp = file_path.open("rb")
         rows, format_used = rows_from_file(fp, format=format, encoding=encoding)
         tracker = None
-        if format_used in (Format.CSV, Format.TSV) and not no_detect_types:
-            tracker = TypeTracker()
-            rows = tracker.wrap(rows)
-        if flatten:
-            rows = (_flatten(row) for row in rows)
-        db[file_table].insert_all(rows, alter=True)
+        try:
+            if format_used in (Format.CSV, Format.TSV) and not no_detect_types:
+                tracker = TypeTracker()
+                rows = tracker.wrap(rows)
+            if flatten:
+                rows = (_flatten(row) for row in rows)
+            db[file_table].insert_all(rows, alter=True)
+        except UnicodeDecodeError as e:
+            fp.close()
+            raise e
         if tracker is not None:
             db[file_table].transform(types=tracker.types)
         # Add convenient t / t1 / t2 views
@@ -3196,8 +3200,12 @@ def __init__(self, exception, path):
     "ctime": lambda p: p.stat().st_ctime,
     "mtime_int": lambda p: int(p.stat().st_mtime),
     "ctime_int": lambda p: int(p.stat().st_ctime),
-    "mtime_iso": lambda p: datetime.utcfromtimestamp(p.stat().st_mtime).isoformat(),
-    "ctime_iso": lambda p: datetime.utcfromtimestamp(p.stat().st_ctime).isoformat(),
+    "mtime_iso": lambda p: datetime.fromtimestamp(p.stat().st_mtime, timezone.utc)
+    .isoformat()
+    .rstrip("+00:00"),
+    "ctime_iso": lambda p: datetime.fromtimestamp(p.stat().st_mtime, timezone.utc)
+    .isoformat()
+    .rstrip("+00:00"),
     "size": lambda p: p.stat().st_size,
     "stem": lambda p: p.stem,
     "suffix": lambda p: p.suffix,
diff --git a/sqlite_utils/utils.py b/sqlite_utils/utils.py
index 9e9882a9..8c010f08 100644
--- a/sqlite_utils/utils.py
+++ b/sqlite_utils/utils.py
@@ -212,6 +212,7 @@ def _extra_key_strategy(
     reader: Iterable[dict],
     ignore_extras: Optional[bool] = False,
     extras_key: Optional[str] = None,
+    buffer: Optional[io.TextIOWrapper] = None,
 ) -> Iterable[dict]:
     # Logic for handling CSV rows with more values than there are headings
     for row in reader:
@@ -231,6 +232,8 @@ def _extra_key_strategy(
         else:
             row[extras_key] = row.pop(None)  # type: ignore
             yield row
+    if buffer:
+        buffer.close()
 
 
 def rows_from_file(
@@ -299,7 +302,10 @@ class Format(enum.Enum):
             reader = csv.DictReader(decoded_fp, dialect=dialect)
         else:
             reader = csv.DictReader(decoded_fp)
-        return _extra_key_strategy(reader, ignore_extras, extras_key), Format.CSV
+        return (
+            _extra_key_strategy(reader, ignore_extras, extras_key, decoded_fp),
+            Format.CSV,
+        )
     elif format == Format.TSV:
         rows = rows_from_file(
             fp, format=Format.CSV, dialect=csv.excel_tab, encoding=encoding