Secure supply chain and continuous delivery

What you will be able to do

1. Inventory code, model, data, dependency, and image artifacts.
2. Verify hashes, provenance, licenses, and scan results.
3. Gate changes on evaluation and rollback evidence.

Prerequisites: Builder-level interface validation, Python, testing, metadata logging and state-transition tracing. Hardware benchmarks and model accounts are not required.

This curriculum runs self-study on VetsLearn. Where the text calls for an instructor, peer or formal review, use the stated acceptance criteria for self-review. Reflection questions support reflection; they do not replace independent assessment. A peer review is optional unless your actual organization requires it.

Part 1 Technical instruction

An AI release contains more than source code. Inventory dependencies, container/base image, model identifier and checksum, prompt/config versions, dataset/eval manifests, licenses, and build provenance. A hash proves byte identity, not trust or authorization. CI classifies the change and runs required gates: unit/security tests, dependency policy, model/data hash approval, evaluation slices, configuration validation, and rollback availability. A signed bundle is useful only when the signer, key custody, verification process, and source inputs are governed.

Mechanism and implementation

A release includes source, dependencies, images, prompts, model/adapter revisions, index manifests and evaluators. Verify provenance/licensing separately from checksums. Mutable aliases can change outside your commit; monitor and reevaluate them. Classify changes to choose appropriate gates.

Constructed offline example

Save as topic_example.py and run python topic_example.py with an installed Python 3.11+ interpreter. This standard-library construction uses synthetic values; its checks are not provider measurements.

import hashlib
artifact = b"synthetic-model-v1"
expected = hashlib.sha256(artifact).hexdigest()
def gate(payload, origin_approved, license_reviewed):
    return hashlib.sha256(payload).hexdigest() == expected and origin_approved and license_reviewed
assert gate(artifact, True, True)
assert not gate(artifact, False, True)
assert not gate(b"changed", True, True)
print("identity plus provenance gates passed")

Engineering decisions

• Define the release artifact set.
• Map change classes to required gates.
• Block unknown provenance or changed critical hashes pending review.

Worked example

The supplied manifest contains one changed model hash and a vulnerable synthetic dependency flag. The pipeline blocks release, records causes, then passes a corrected, independently reproducible bundle.

Diagnose the failure

Hashes establish byte identity only. Unknown origins remain quarantined. Inventory every runtime-loaded artifact before relying on signed bundles.

Part 2 Guided practice

Predict why unchanged unapproved bytes fail. Make a release manifest including prompt/tool/eval revisions. Independently change only the prompt and map it to required behavior/security tests.

Answer criteria and feedback: Hashes establish byte identity only. Unknown origins remain quarantined. Inventory every runtime-loaded artifact before relying on signed bundles.

Continue with the downloadable pack below. If the pack or interpreter is unavailable, trace this self-contained example by hand in an input/expected-status/controlling-rule table and label execution unverified. Do not invent passing output. No account, credential, paid call, purchase or actual deployment is required.

Part 3 Remediation

Verify one artifact hash manually. If provenance is unknown, quarantine the artifact rather than adding it to the allowlist. After repair, reassess with the alternate scenario in scenario_alt.json. A critical authorization, data-handling, evidence-integrity, or prohibited-action failure cannot be averaged into successful practice; it requires targeted remediation.

Part 4 Sources and currency

• https://csrc.nist.gov/pubs/sp/800/218/a/final
• https://www.cisa.gov/sbom

Use the workflow decision reference for model/deployment/harness distinctions and controlled comparisons. Source currency must be checked before each cohort. Provider/product behavior is illustrative where mocked; resolve current interfaces only from official documentation in an authorized environment.