diff --git a/api/onnx_web/chain/edit_safety.py b/api/onnx_web/chain/edit_safety.py
new file mode 100644
index 00000000..3f3c038b
--- /dev/null
+++ b/api/onnx_web/chain/edit_safety.py
@@ -0,0 +1,72 @@
+from logging import getLogger
+from typing import Optional
+
+from PIL import Image
+
+from ..params import ImageParams, SizeChart, StageParams
+from ..server import ServerContext
+from ..worker import ProgressCallback, WorkerContext
+from .base import BaseStage
+from .result import StageResult
+
+logger = getLogger(__name__)
+
+
+class EditSafetyStage(BaseStage):
+    max_tile = SizeChart.max
+
+    def run(
+        self,
+        _worker: WorkerContext,
+        server: ServerContext,
+        _stage: StageParams,
+        _params: ImageParams,
+        sources: StageResult,
+        *,
+        callback: Optional[ProgressCallback] = None,
+        **kwargs,
+    ) -> StageResult:
+        logger.info("checking results using horde safety")
+
+        # keep these within run to make this sort of like a plugin or peer dependency
+        try:
+            from horde_safety.deep_danbooru_model import get_deep_danbooru_model
+            from horde_safety.interrogate import get_interrogator_no_blip
+            from horde_safety.nsfw_checker_class import NSFWChecker
+
+            # set up
+            block_nsfw = server.has_feature("horde-safety-nsfw")
+
+            interrogator = get_interrogator_no_blip()
+            deep_danbooru_model = get_deep_danbooru_model()
+
+            nsfw_checker = NSFWChecker(
+                interrogator,
+                deep_danbooru_model,
+            )
+
+            # individual flags from NSFWResult
+            is_csam = False
+
+            images = sources.as_images()
+            results = []
+            for i, image in enumerate(images):
+                prompt = sources.metadata[i].prompt
+                check = nsfw_checker.check_for_nsfw(image, prompt=prompt)
+
+                if check.is_csam:
+                    logger.warning("flagging csam result: %s, %s", i, prompt)
+                    is_csam = True
+
+                if check.is_nsfw and block_nsfw:
+                    logger.warning("blocking nsfw image: %s, %s", i, prompt)
+                    results.append(Image.new("RGB", image.size, color="black"))
+
+            if is_csam:
+                logger.warning("blocking csam result")
+                raise RuntimeError("csam detected")
+            else:
+                return StageResult.from_images(results, metadata=sources.metadata)
+        except ImportError:
+            logger.warning("horde safety not installed")
+            return StageResult.empty()
diff --git a/api/onnx_web/chain/stages.py b/api/onnx_web/chain/stages.py
index 1e094052..9ad44613 100644
--- a/api/onnx_web/chain/stages.py
+++ b/api/onnx_web/chain/stages.py
@@ -1,5 +1,8 @@
 from logging import getLogger
 
+from .edit_safety import EditSafetyStage
+from .edit_text import EditTextStage
+
 from .base import BaseStage
 from .blend_denoise_fastnlmeans import BlendDenoiseFastNLMeansStage
 from .blend_denoise_localstd import BlendDenoiseLocalStdStage
@@ -40,6 +43,8 @@ CHAIN_STAGES = {
     "correct-codeformer": CorrectCodeformerStage,
     "correct-gfpgan": CorrectGFPGANStage,
     "edit-metadata": EditMetadataStage,
+    "edit-safety": EditSafetyStage,
+    "edit-text": EditTextStage,
     "persist-disk": PersistDiskStage,
     "persist-s3": PersistS3Stage,
     "reduce-crop": ReduceCropStage,
diff --git a/api/onnx_web/diffusers/run.py b/api/onnx_web/diffusers/run.py
index bb90e0ab..971c6b0c 100644
--- a/api/onnx_web/diffusers/run.py
+++ b/api/onnx_web/diffusers/run.py
@@ -52,6 +52,16 @@ def get_highres_tile(
     return params.unet_tile
 
 
+def add_safety_stage(
+    server: ServerContext,
+    pipeline: ChainPipeline,
+) -> None:
+    if server.has_feature("horde-safety"):
+        from ..chain.edit_safety import EditSafetyStage
+
+        pipeline.stage(EditSafetyStage(), StageParams())
+
+
 def run_txt2img_pipeline(
     worker: WorkerContext,
     server: ServerContext,
@@ -110,6 +120,8 @@ def run_txt2img_pipeline(
         upscale=after_upscale,
     )
 
+    add_safety_stage(server, chain)
+
     # run and save
     latents = get_latents_from_seed(params.seed, size, batch=params.batch)
     progress = worker.get_progress_callback(reset=True)
@@ -209,6 +221,8 @@ def run_img2img_pipeline(
         chain=chain,
     )
 
+    add_safety_stage(server, chain)
+
     # run and append the filtered source
     progress = worker.get_progress_callback(reset=True)
     images = chain(
@@ -382,6 +396,8 @@ def run_inpaint_pipeline(
         chain=chain,
     )
 
+    add_safety_stage(server, chain)
+
     # run and save
     latents = get_latents_from_seed(params.seed, size, batch=params.batch)
     progress = worker.get_progress_callback(reset=True)
@@ -462,6 +478,8 @@ def run_upscale_pipeline(
         chain=chain,
     )
 
+    add_safety_stage(server, chain)
+
     # run and save
     progress = worker.get_progress_callback(reset=True)
     images = chain(
@@ -515,6 +533,8 @@ def run_blend_pipeline(
         chain=chain,
     )
 
+    add_safety_stage(server, chain)
+
     # run and save
     progress = worker.get_progress_callback(reset=True)
     images = chain(