microsoft · rlundeen2 · Jun 22, 2026 · Jun 22, 2026 · Jun 23, 2026 · Jun 23, 2026
diff --git a/.pyrit_conf_example b/.pyrit_conf_example
@@ -26,7 +26,8 @@ memory_db_type: sqlite
 # - airt: AI Red Team setup with Azure OpenAI (requires AZURE_OPENAI_* env vars)
 # - target: Registers available prompt targets into the TargetRegistry
 # - scorer: Registers pre-configured scorers into the ScorerRegistry
-# - load_default_datasets: Loads default datasets for all registered scenarios
+# - load_default_datasets: Optional preload of default datasets for all registered
+#   scenarios (scenarios otherwise fetch their datasets on demand)
 # - objective_list: Sets default objectives for scenarios
 #
 # Each initializer can be specified as:
@@ -46,7 +47,6 @@ memory_db_type: sqlite
 #         - scorer
 initializers:
   - name: simple
-  - name: load_default_datasets
   - name: target
     args:
       tags:

diff --git a/doc/getting_started/pyrit_conf.md b/doc/getting_started/pyrit_conf.md
@@ -111,7 +111,6 @@ Most users should enable the following initializers. These are what the `.pyrit_
 | `simple` | Baseline defaults for converters, scorers, and attack configs using your `OPENAI_CHAT_*` env vars | Always — provides the foundation for most PyRIT operations |
 | `target` | Prompt targets (OpenAI, Azure, AML, etc.) into the `TargetRegistry` | **Required for `pyrit_scan`** and any registry-based workflows |
 | `scorer` | Scorers (refusal, content safety, harm-category, Likert, etc.) into the `ScorerRegistry` | **Required for automated scoring** and `pyrit_scan` evaluations |
-| `load_default_datasets` | Seed datasets for all registered scenarios into memory | **Required for `pyrit_scan` scenarios** — they need data to run |
 
 ```{note}
 **Execution order follows listing order.** Initializers execute in the order they appear in the config. Ensure dependencies are satisfied — for example, list `target` before `scorer` since scorers need targets to be registered first.
@@ -122,7 +121,6 @@ The recommended config:
 ```yaml
 initializers:
   - name: simple
-  - name: load_default_datasets
   - name: scorer
   - name: target
     args:
@@ -131,6 +129,10 @@ initializers:
         - scorer
 ```
 
+```{note}
+**`load_default_datasets` is optional.** Scenarios fetch their datasets from the registered provider on demand the first time they run, so you no longer need this initializer for everyday runs. Add it only when you want to preload every scenario's datasets up front — for example, to warm memory for repeated runs or to populate a database for offline use.
+```
+
 ### `initialization_scripts`
 
 Paths to custom Python scripts containing `PyRITInitializer` subclasses. Paths can be absolute or relative to the current working directory.

diff --git a/doc/index.md b/doc/index.md
@@ -101,7 +101,6 @@ initializers:
         - default
         - scorer
   - name: scorer
-  - name: load_default_datasets
 ```
 ::::
 

diff --git a/doc/scanner/0_scanner.md b/doc/scanner/0_scanner.md
@@ -23,7 +23,7 @@ PyRIT provides two command-line interfaces:
 
 ```bash
 # Run the Foundry RedTeamAgent scenario against your configured target
-pyrit_scan foundry.red_team_agent --target openai_chat --initializers target load_default_datasets --strategies base64
+pyrit_scan foundry.red_team_agent --target openai_chat --initializers target --strategies base64
 ```
 
 ## Built-in Scenarios

diff --git a/doc/scanner/1_pyrit_scan.ipynb b/doc/scanner/1_pyrit_scan.ipynb
@@ -110,7 +110,7 @@
     "Or concretely:\n",
     "\n",
     "```shell\n",
-    "!pyrit_scan foundry.red_team_agent --target openai_chat --initializers load_default_datasets target --scenario-strategies base64\n",
+    "!pyrit_scan foundry.red_team_agent --target openai_chat --initializers target --scenario-strategies base64\n",
     "```\n",
     "\n",
     "Example with a basic configuration that runs the Foundry scenario against the objective target defined in the `target` initializer."
@@ -123,28 +123,28 @@
    "metadata": {},
    "outputs": [],
    "source": [
-    "!pyrit_scan foundry.red_team_agent --target openai_chat --initializers load_default_datasets target --strategies base64"
+    "!pyrit_scan foundry.red_team_agent --target openai_chat --initializers target --strategies base64"
    ]
   },
   {
    "cell_type": "markdown",
    "id": "8",
    "metadata": {},
    "source": [
-    "Or with all options and multiple initializers and multiple strategies:\n",
+    "Or with all options and multiple strategies:\n",
     "\n",
     "```shell\n",
-    "pyrit_scan foundry.red_team_agent --target openai_chat --initializers load_default_datasets target --strategies easy crescendo\n",
+    "pyrit_scan foundry.red_team_agent --target openai_chat --initializers target --strategies easy crescendo\n",
     "```\n",
     "\n",
     "You can also override scenario execution parameters:\n",
     "\n",
     "```shell\n",
     "# Override concurrency and retry settings\n",
-    "pyrit_scan foundry.red_team_agent --target openai_chat --initializers load_default_datasets target --max-concurrency 10 --max-retries 3\n",
+    "pyrit_scan foundry.red_team_agent --target openai_chat --initializers target --max-concurrency 10 --max-retries 3\n",
     "\n",
     "# Add custom memory labels for tracking (must be valid JSON)\n",
-    "pyrit_scan foundry.red_team_agent --target openai_chat --initializers load_default_datasets target --memory-labels '{\"experiment\": \"test1\", \"version\": \"v2\", \"researcher\": \"alice\"}'\n",
+    "pyrit_scan foundry.red_team_agent --target openai_chat --initializers target --memory-labels '{\"experiment\": \"test1\", \"version\": \"v2\", \"researcher\": \"alice\"}'\n",
     "```\n",
     "\n",
     "Available CLI parameter overrides:\n",

diff --git a/doc/scanner/1_pyrit_scan.py b/doc/scanner/1_pyrit_scan.py
@@ -78,29 +78,29 @@
 # Or concretely:
 #
 # ```shell
-# !pyrit_scan foundry.red_team_agent --target openai_chat --initializers load_default_datasets target --scenario-strategies base64
+# !pyrit_scan foundry.red_team_agent --target openai_chat --initializers target --scenario-strategies base64
 # ```
 #
 # Example with a basic configuration that runs the Foundry scenario against the objective target defined in the `target` initializer.
 
 # %%
-# !pyrit_scan foundry.red_team_agent --target openai_chat --initializers load_default_datasets target --strategies base64
+# !pyrit_scan foundry.red_team_agent --target openai_chat --initializers target --strategies base64
 
 # %% [markdown]
-# Or with all options and multiple initializers and multiple strategies:
+# Or with all options and multiple strategies:
 #
 # ```shell
-# pyrit_scan foundry.red_team_agent --target openai_chat --initializers load_default_datasets target --strategies easy crescendo
+# pyrit_scan foundry.red_team_agent --target openai_chat --initializers target --strategies easy crescendo
 # ```
 #
 # You can also override scenario execution parameters:
 #
 # ```shell
 # # Override concurrency and retry settings
-# pyrit_scan foundry.red_team_agent --target openai_chat --initializers load_default_datasets target --max-concurrency 10 --max-retries 3
+# pyrit_scan foundry.red_team_agent --target openai_chat --initializers target --max-concurrency 10 --max-retries 3
 #
 # # Add custom memory labels for tracking (must be valid JSON)
-# pyrit_scan foundry.red_team_agent --target openai_chat --initializers load_default_datasets target --memory-labels '{"experiment": "test1", "version": "v2", "researcher": "alice"}'
+# pyrit_scan foundry.red_team_agent --target openai_chat --initializers target --memory-labels '{"experiment": "test1", "version": "v2", "researcher": "alice"}'
 # ```
 #
 # Available CLI parameter overrides:

diff --git a/doc/scanner/2_pyrit_shell.md b/doc/scanner/2_pyrit_shell.md
@@ -25,7 +25,7 @@ pyrit_shell --config-file ./.pyrit_conf
 pyrit_shell --log-level DEBUG
 
 # Load initializers at startup
-pyrit_shell --initializers load_default_datasets
+pyrit_shell --initializers target
 
 # Load custom initialization scripts
 pyrit_shell --initialization-scripts ./my_config.py
@@ -54,32 +54,32 @@ The `run` command executes scenarios with the same options as `pyrit_scan`:
 ### Basic Usage
 
 ```bash
-pyrit> run foundry.red_team_agent --target my_target --initializers target load_default_datasets
+pyrit> run foundry.red_team_agent --target my_target --initializers target
 ```
 
 ### With Strategies
 
 ```bash
-pyrit> run garak.encoding --target my_target --initializers target load_default_datasets --strategies base64 rot13
+pyrit> run garak.encoding --target my_target --initializers target --strategies base64 rot13
 
-pyrit> run foundry.red_team_agent --target my_target --initializers target load_default_datasets -s jailbreak crescendo
+pyrit> run foundry.red_team_agent --target my_target --initializers target -s jailbreak crescendo
 ```
 
 ### With Runtime Parameters
 
 ```bash
 # Set concurrency and retries
-pyrit> run foundry.red_team_agent --target my_target --initializers target load_default_datasets --max-concurrency 10 --max-retries 3
+pyrit> run foundry.red_team_agent --target my_target --initializers target --max-concurrency 10 --max-retries 3
 
 # Add memory labels for tracking
-pyrit> run garak.encoding --target my_target --initializers target load_default_datasets --memory-labels '{"experiment":"test1","version":"v2"}'
+pyrit> run garak.encoding --target my_target --initializers target --memory-labels '{"experiment":"test1","version":"v2"}'
 ```
 
 ### Override Defaults Per-Run
 
 ```bash
 # Override log level for this run only
-pyrit> run garak.encoding --target my_target --initializers target load_default_datasets --log-level DEBUG
+pyrit> run garak.encoding --target my_target --initializers target --log-level DEBUG
 ```
 
 ### Run Command Options
@@ -119,9 +119,9 @@ pyrit> scenario-history
 
 Scenario Run History:
 ================================================================================
-1) foundry.red_team_agent --initializers target load_default_datasets --strategies base64
-2) garak.encoding --initializers target load_default_datasets --strategies rot13
-3) foundry.red_team_agent --initializers target load_default_datasets -s jailbreak
+1) foundry.red_team_agent --initializers target --strategies base64
+2) garak.encoding --initializers target --strategies rot13
+3) foundry.red_team_agent --initializers target -s jailbreak
 ================================================================================
 
 Total runs: 3
@@ -135,7 +135,7 @@ The shell excels at interactive testing workflows:
 
 ```bash
 # Start shell with defaults
-pyrit_shell --initializers target load_default_datasets
+pyrit_shell --initializers target
 
 # Quick exploration
 pyrit> list-scenarios
@@ -166,7 +166,7 @@ pyrit> print-scenario 2
 
 2. **Use short strategy aliases** with `-s`:
    ```bash
-   pyrit> run foundry.red_team_agent --initializers target load_default_datasets -s base64 rot13
+   pyrit> run foundry.red_team_agent --initializers target -s base64 rot13
    ```
 
 3. **Review history regularly** to track what you've tested:

diff --git a/doc/scanner/airt.ipynb b/doc/scanner/airt.ipynb
@@ -72,7 +72,7 @@
     "\n",
     "```bash\n",
     "pyrit_scan airt.rapid_response \\\n",
-    "  --initializers target load_default_datasets \\\n",
+    "  --initializers target \\\n",
     "  --target openai_chat \\\n",
     "  --strategies role_play \\\n",
     "  --dataset-names airt_hate \\  \n",
@@ -367,7 +367,7 @@
     "\n",
     "```bash\n",
     "pyrit_scan airt.cyber \\\n",
-    "  --initializers target load_default_datasets \\\n",
+    "  --initializers target \\\n",
     "  --target openai_chat \\\n",
     "  --strategies multi_turn \\\n",
     "  --max-dataset-size 1\n",
@@ -504,7 +504,7 @@
     "\n",
     "```bash\n",
     "pyrit_scan airt.jailbreak \\\n",
-    "  --initializers target load_default_datasets \\\n",
+    "  --initializers target \\\n",
     "  --target openai_chat \\\n",
     "  --strategies prompt_sending \\\n",
     "  --max-dataset-size 1\n",
@@ -1125,7 +1125,7 @@
     "\n",
     "```bash\n",
     "pyrit_scan airt.scam \\\n",
-    "  --initializers target load_default_datasets \\\n",
+    "  --initializers target \\\n",
     "  --target openai_chat \\\n",
     "  --strategies context_compliance \\\n",
     "  --max-dataset-size 1\n",

diff --git a/doc/scanner/airt.py b/doc/scanner/airt.py
@@ -40,7 +40,7 @@
 #
 # ```bash
 # pyrit_scan airt.rapid_response \
-#   --initializers target load_default_datasets \
+#   --initializers target \
 #   --target openai_chat \
 #   --strategies role_play \
 #   --dataset-names airt_hate \
@@ -123,7 +123,7 @@
 #
 # ```bash
 # pyrit_scan airt.cyber \
-#   --initializers target load_default_datasets \
+#   --initializers target \
 #   --target openai_chat \
 #   --strategies multi_turn \
 #   --max-dataset-size 1
@@ -156,7 +156,7 @@
 #
 # ```bash
 # pyrit_scan airt.jailbreak \
-#   --initializers target load_default_datasets \
+#   --initializers target \
 #   --target openai_chat \
 #   --strategies prompt_sending \
 #   --max-dataset-size 1
@@ -236,7 +236,7 @@
 #
 # ```bash
 # pyrit_scan airt.scam \
-#   --initializers target load_default_datasets \
+#   --initializers target \
 #   --target openai_chat \
 #   --strategies context_compliance \
 #   --max-dataset-size 1

diff --git a/doc/scanner/benchmark.ipynb b/doc/scanner/benchmark.ipynb
@@ -31,7 +31,7 @@
     "\n",
     "```bash\n",
     "pyrit_scan benchmark.adversarial \\\n",
-    "  --initializers target load_default_datasets \\\n",
+    "  --initializers target \\\n",
     "  --target openai_chat \\\n",
     "  --adversarial-targets adversarial_chat_singleturn adversarial_chat_multiturn \\\n",
     "  --max-dataset-size 4\n",

diff --git a/doc/scanner/benchmark.py b/doc/scanner/benchmark.py
@@ -30,7 +30,7 @@
 #
 # ```bash
 # pyrit_scan benchmark.adversarial \
-#   --initializers target load_default_datasets \
+#   --initializers target \
 #   --target openai_chat \
 #   --adversarial-targets adversarial_chat_singleturn adversarial_chat_multiturn \
 #   --max-dataset-size 4

diff --git a/doc/scanner/pyrit_conf.yaml b/doc/scanner/pyrit_conf.yaml
@@ -8,4 +8,3 @@ initializers:
         - scorer
   - name: scorer
   - name: scenario_technique
-  - name: load_default_datasets
diff --git a/pyrit/backend/services/scenario_run_service.py b/pyrit/backend/services/scenario_run_service.py
@@ -24,7 +24,7 @@
 from pyrit.models import AttackOutcome, ScenarioResult
 from pyrit.registry import InitializerRegistry, ScenarioRegistry, TargetRegistry
 from pyrit.scenario import Scenario
-from pyrit.scenario.core import DatasetConfiguration
+from pyrit.scenario.core import DatasetAttackConfiguration
 
 if TYPE_CHECKING:
     from pyrit.prompt_target import PromptTarget
@@ -266,11 +266,11 @@ def _build_init_kwargs(
         Resolves strategies and dataset configuration from the request.
 
         Dataset configuration is built so that the scenario's default
-        ``DatasetConfiguration`` *subclass* (e.g. ``EncodingDatasetConfiguration``)
+        ``DatasetAttackConfiguration`` *subclass* (e.g. ``EncodingDatasetConfiguration``)
         is preserved when the caller overrides ``dataset_names`` or
         ``max_dataset_size``. Subclasses commonly override
-        ``get_all_seed_attack_groups()`` or ``_load_seed_groups_for_dataset()``
-        to shape seeds into scenario-appropriate ``SeedAttackGroup`` objects.
+        ``_build_attack_groups()`` to shape seeds into scenario-appropriate
+        ``SeedAttackGroup`` objects.
 
         Args:
             request: The run request.
@@ -345,18 +345,18 @@ def _build_init_kwargs(
                 except TypeError as exc:
                     # The subclass __init__ takes extra required kwargs we cannot
                     # supply from a backend request. Fall back to the base
-                    # DatasetConfiguration so the run can still proceed; downstream
+                    # DatasetAttackConfiguration so the run can still proceed; downstream
                     # scenarios that strictly require the subclass should either
                     # define a no-extra-required-args constructor or surface the
                     # incompatibility through their own initialize_async validation.
                     logger.warning(
                         "Cannot construct %s(dataset_names=..., max_dataset_size=...) (%s). "
-                        "Falling back to a generic DatasetConfiguration; scenario-specific "
+                        "Falling back to a generic DatasetAttackConfiguration; scenario-specific "
                         "dataset-config behavior may be lost.",
                         default_config_class.__name__,
                         exc,
                     )
-                    init_kwargs["dataset_config"] = DatasetConfiguration(
+                    init_kwargs["dataset_config"] = DatasetAttackConfiguration(
                         dataset_names=request.dataset_names,
                         max_dataset_size=request.max_dataset_size,
                     )

diff --git a/pyrit/models/__init__.py b/pyrit/models/__init__.py
@@ -107,6 +107,7 @@
     SeedSimulatedConversation,
     SeedUnion,
     SimulatedTargetSystemPromptPaths,
+    group_seeds_into_attack_groups,
 )
 from pyrit.models.target_capabilities import CapabilityName, TargetCapabilities
 
@@ -153,6 +154,7 @@
     "get_all_values",
     "group_conversation_message_pieces_by_sequence",
     "group_message_pieces_into_conversations",
+    "group_seeds_into_attack_groups",
     "HarmDefinition",
     "Identifiable",
     "IdentifierFilter",

diff --git a/pyrit/models/seeds/__init__.py b/pyrit/models/seeds/__init__.py
@@ -20,6 +20,7 @@
 from pyrit.models.seeds.seed_attack_technique_group import SeedAttackTechniqueGroup
 from pyrit.models.seeds.seed_dataset import SeedDataset
 from pyrit.models.seeds.seed_group import SeedGroup, SeedUnion
+from pyrit.models.seeds.seed_grouping import group_seeds_into_attack_groups
 from pyrit.models.seeds.seed_objective import SeedObjective
 from pyrit.models.seeds.seed_prompt import SeedPrompt
 from pyrit.models.seeds.seed_simulated_conversation import (
@@ -37,6 +38,7 @@
     "load_seed_dataset_from_yaml",
     "load_seed_from_yaml",
     "load_seed_prompt_from_yaml_with_required_parameters",
+    "group_seeds_into_attack_groups",
     "NextMessageSystemPromptPaths",
     "Seed",
     "SeedAttackGroup",
-Original file line number
+Diff line change
@@ Expand Up / @@ -101,7 +101,6 @@ initializers: @@
             - default
             - scorer
       - name: scorer
-      - name: load_default_datasets
     ```
     ::::
@@ Expand Down @@