Use yaml.safe_load by default in Python

2026-04-15

yaml.load with unsafe loaders can construct Python objects from YAML and trigger code execution. Use yaml.safe_load unless you explicitly need trusted custom object deserialization.

Goal

This post explains why yaml.safe_load should be your default in Python.

The short reason is simple: yaml.load with unsafe loaders can execute behavior from YAML input.

The Risk in One Example

import yaml

yaml_text = """
!!python/object/apply:builtins.print
- "Hello from yaml.load"
"""

print("Before load")

yaml.load(yaml_text, Loader=yaml.Loader)

print("After load")

If you run it, the print inside YAML is executed during deserialization.

Expected output:

Before load
Hello from yaml.load
After load

This is exactly the problem. The parser is not only reading data. It is also interpreting Python-specific tags.

Why This Happens

yaml.Loader supports Python object construction tags such as:

  • !!python/object
  • !!python/object/new
  • !!python/object/apply

That feature can be useful in very narrow trusted environments, but it is dangerous for general config/input parsing.

If YAML comes from users, APIs, files you do not fully control, or any mixed-trust boundary, this can become a code execution path.

Safe Alternative

Use yaml.safe_load:

import yaml

data = yaml.safe_load(yaml_text)
print(data)

safe_load only parses standard YAML types (mapping, list, string, number, bool, null, etc.). It does not allow Python object construction tags.

So malicious payloads like !!python/object/apply:... are not executed.

Takeaway

yaml.load can deserialize more than data.

If your goal is to parse configuration safely, yaml.safe_load is the right default.