NCA-GENM domain - 5% of the exam

Trustworthy AI

Trustworthy AI is 5% of the NVIDIA-Certified Associate: Generative AI Multimodal (NCA-GENM) exam. These are the objectives it covers, each with practice questions and worked explanations.

Objectives in this domain

Sample question from this domain

Free sampleTrustworthy AImedium

A media organisation wants to attach tamper-evident provenance data to every AI-generated image before publishing. Which open standard is designed specifically to carry cryptographically signed assertions about a content item's origin and editing history?

  • AC2PA (Coalition for Content Provenance and Authenticity) Content Credentials Correct
  • BEXIF metadata embedded in the image file header
  • CCLIP embeddings stored alongside the image as a sidecar file
  • DA perceptual hash digest appended to the filename
Identify the C2PA standard as the mechanism for attaching cryptographically signed provenance data to AI-generated content. The Coalition for Content Provenance and Authenticity (C2PA) specification defines Content Credentials: a signed manifest that binds assertions - such as generator identity, model used, and editing actions - to a content item using public-key cryptography. Because the manifest is signed, any tampering with either the content or its assertions can be detected by a verifier. This is the only option among the four that provides both tamper evidence and a portable, standardised chain of custody.

Why A is correct: C2PA defines a cryptographically signed manifest that records assertions about how content was created or modified, enabling downstream verification of origin and editing history.

Why B is wrong: EXIF stores camera and device data but carries no cryptographic signatures, so it can be stripped or forged without detection and does not constitute a provenance standard.

Why C is wrong: CLIP produces semantic similarity embeddings for retrieval and classification tasks; it has no signing mechanism and cannot provide tamper-evident provenance assertions.

Why D is wrong: Perceptual hashes measure visual similarity and can detect near-duplicates, but they carry no chain-of-custody record and are not signed, so they cannot prove origin or detect editorial changes.

Other domains in this exam

See also the NCA-GENM cert hub, the study guide, and the cheat sheet.

Examworthy is not affiliated with or endorsed by NVIDIA. Original, blueprint-aligned practice material only.