Verifiable Inference
Cryptographic proof that a claimed LLM output was actually produced by a specified model on a specified input, without needing to trust the provider. Uses zero-knowledge proofs or trusted-execution environments. Emerging in 2026 for regulated industries where output provenance is auditable.
Last reviewed:
Related terms
- Watermarking (AI content)
Embedding a statistically-detectable signal into generated content so downstream verifiers can identify it as AI-produced. Different implementations for text (token-level bias) and images (imperceptible frequency-domain patterns). Increasingly part of provider defaults; complements C2PA content provenance.
- C2PA
Coalition for Content Provenance and Authenticity — an open standard for cryptographically signing the origin and edit history of digital content (images, video, audio, PDF). Adopted by Adobe, Microsoft, OpenAI, Google, TruePic. Foundational for AI-generated content labelling under the EU AI Act.
- Groundedness
A measure of how well an LLM output is supported by provided source material. High-groundedness answers cite or paraphrase provided context; low-groundedness answers introduce claims not present in sources. A primary target metric for RAG systems.
- AI Governance
The framework of policies, procedures, and controls that guide the development, deployment, and use of AI systems within an organization. Ensures AI aligns with organizational values and regulations.
Related on this site
Where this fits
Verifiable Inference is part of the Evaluation & Governance vocabulary used in the Generative AI Maturity Framework. See the full glossary for the complete set of 159 defined terms, or take the free maturity assessment to see where your organisation stands.