liveBook · Manning

Traceability

As explained in AI Governance

Traceability is the ability to identify and follow the models, versions, inputs, and evidence that contributed to an outcome.

Traceability means being able to identify which model, version, prompts, and evidence shaped an answer. It makes system behavior more accountable when automated decisions affect people. section 5.5.3

Traceability does not require access to model weights or source code. It does require records that support explanations and disclosure; without the ability to explain, trace, or disclose, an organization cannot claim compliance. section 5.5.3

Transparency controls can include:

  • Notifying job candidates when an automated employment decision tool is used, identifying the qualifications or characteristics it assesses, and publishing annual bias audits. section 5.5.3
  • Publishing transparency reports and disclosures for automated decision systems used in public administration. section 5.5.3

How AI Governance covers it

How deep it goes
Defines expected trace records and distinguishes them from model internals, while deferring implementation methods such as logging and structured prompts.
Examples
Uses employment-tool notices and audits, plus transparency reports for public-sector automated decisions.
What you can do after
Evaluate whether records identify the model, version, prompts, and evidence behind an answer.
How it connects
Connects traceability with disclosure, explanations, automated decision-making, and bias audits.

Also explained in

Compare how 4 books explain traceability →

Related concepts

This explainer was generated by Manning's knowledge graph from the text of AI Governance. The section links point to the parts of the book it draws on.