liveBook · Manning

GRC program

As explained in AI Governance

An organizational program that integrates governance, risk management, and compliance activities for a technology or business area.

A governance, risk management, and compliance (GRC) program coordinates how an organization identifies, controls, and responds to risks. For generative AI, it connects governance decisions to engineering safeguards and operational monitoring. It cannot eliminate hallucinations or false positives, but can help detect and correct them before they escalate. section 1.1

Document vetting under attack

A vendor uses an open-source LLM to check passport photos and government IDs for an airline. Hidden instructions in an uploaded image trick the model into bypassing checks, approving invalid documents and creating security, regulatory, and reputational consequences. section 1.1

Controls should span deployment and operation:

  • Before deployment, test safety and bias, run adversarial prompt suites, and red-team the system. section 1.1
  • After launch, use output filters, rate limits, live toxicity and drift metrics, and anomaly detection. section 1.1
Table 1.1 Why Generative AI Breaks Traditional Control Models (Continued)
Pre-deployment assurance
Tests, code review, static analysis, red-teaming (ethical hackers)
Role-specific training, competency certification, scenario-based drills
Safety and bias testing, adversarial prompt suites, red-teaming, interpretability tools (e.g. probing)
Post-launch controls
Logging, health checks, intrusion detection, WAF, auto-rollback on error rate spike
Supervisors, peer review, escalation protocols
Output filters, rate limiting, live metrics on toxicity and drift, anomaly detection on embeddings
Failure detection & attribution
Clear (stack traces, logs, input replay)
Mixed (relies on human reporting  or observation)
Opaque (failures can appear valid; root cause buried in model weights)
Memory / learning capability
None (unless explicitly coded)
Present (experience, judgment, learning over time)
Present (training data, fine-tuning, context memory)

Compare pre-deployment assurance, post-launch controls, and failure attribution across software, human operators, and generative AI. section 1.1

How AI Governance covers it

How deep it goes
Explains the rationale and compares deployment controls, while deferring detailed hallucination-containment techniques.
Examples
Uses an airline document-vetting attack and a healthcare documentation enforcement action to illustrate consequences.
What you can do after
Evaluate/understand generative AI risks and compare pre-deployment assurance with post-launch controls.
How it connects
Links hallucinations and prompt injection to adversarial testing, monitoring, and human review.

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.