diff --git a/elk/run.py b/elk/run.py
index fb8903cc..a3ab2b95 100644
--- a/elk/run.py
+++ b/elk/run.py
@@ -1,5 +1,6 @@
 import os
 import random
+import subprocess
 from abc import ABC, abstractmethod
 from collections import defaultdict
 from dataclasses import dataclass
@@ -17,6 +18,8 @@
 from torch import Tensor
 from tqdm import tqdm
 
+import elk
+
 from .debug_logging import save_debug_log
 from .extraction import Extract, extract
 from .extraction.dataset_name import DatasetDictWithName
@@ -31,6 +34,19 @@
 )
 
 
+def fetch_git_hash() -> str | None:
+    try:
+        return (
+            subprocess.check_output(
+                ["git", "rev-parse", "HEAD"], cwd=Path(elk.__file__).parent.parent
+            )
+            .decode("ascii")
+            .strip()
+        )
+    except (NotADirectoryError, subprocess.CalledProcessError):
+        return
+
+
 @dataclass
 class Run(ABC, Serializable):
     data: Extract
@@ -86,15 +102,7 @@ def execute(
         # properly without this flag enabled.
         save(self, self.out_dir / "cfg.yaml", save_dc_types=True)
 
-        path = self.out_dir / "fingerprints.yaml"
-        with open(path, "w") as meta_f:
-            yaml.dump(
-                {
-                    ds_name: {split: ds[split]._fingerprint for split in ds.keys()}
-                    for ds_name, ds in self.datasets
-                },
-                meta_f,
-            )
+        self.write_metadata()
 
         devices = select_usable_devices(self.num_gpus, min_memory=self.min_gpu_mem)
         num_devices = len(devices)
@@ -190,3 +198,21 @@ def apply_to_layers(
                     df.round(4).to_csv(self.out_dir / f"{name}.csv", index=False)
                 if self.debug:
                     save_debug_log(self.datasets, self.out_dir)
+
+    def write_metadata(self):
+        """Write metadata about the run to a yaml file."""
+        assert self.out_dir is not None
+        with open(self.out_dir / "metadata.yaml", "w") as meta_f:
+            dataset_fingerprints = {
+                ds_name: {split: ds[split]._fingerprint for split in ds.keys()}
+                for ds_name, ds in self.datasets
+            }
+            metadata = dict()
+            metadata["datasets"] = dataset_fingerprints
+            git_hash = fetch_git_hash()
+            if git_hash is not None:
+                metadata["git_hash"] = git_hash
+            yaml.dump(
+                metadata,
+                meta_f,
+            )
diff --git a/tests/test_smoke_elicit.py b/tests/test_smoke_elicit.py
index bac0f398..df2d303d 100644
--- a/tests/test_smoke_elicit.py
+++ b/tests/test_smoke_elicit.py
@@ -27,7 +27,7 @@ def test_smoke_elicit_run_tiny_gpt2_ccs(tmp_path: Path):
     created_file_names = {file.name for file in files}
     expected_files = [
         "cfg.yaml",
-        "fingerprints.yaml",
+        "metadata.yaml",
         "lr_models",
         "reporters",
         "eval.csv",
@@ -58,7 +58,7 @@ def test_smoke_elicit_run_tiny_gpt2_eigen(tmp_path: Path):
     created_file_names = {file.name for file in files}
     expected_files = [
         "cfg.yaml",
-        "fingerprints.yaml",
+        "metadata.yaml",
         "lr_models",
         "reporters",
         "eval.csv",
diff --git a/tests/test_smoke_eval.py b/tests/test_smoke_eval.py
index 4efd7112..5d5b74c5 100644
--- a/tests/test_smoke_eval.py
+++ b/tests/test_smoke_eval.py
@@ -9,7 +9,7 @@
 
 EVAL_EXPECTED_FILES = [
     "cfg.yaml",
-    "fingerprints.yaml",
+    "metadata.yaml",
     "eval.csv",
 ]