Topic guide

AI governance

Good AI governance gives people enough visibility, authority, and evidence to make responsible decisions while technology and regulation keep moving. It joins board accountability to the details of data, models, agents, suppliers, and real-world consequences.

Govern the system around the model

Many failures attributed to AI are failures of ownership, incentives, escalation, procurement, or operational discipline. Governance becomes useful when it makes those surrounding conditions explicit and connects them to technical evidence.

Accountability

Name the people who can approve, limit, pause, and retire a system, and ensure they have enough information to act.

Risk appetite

Translate broad values into acceptable uses, prohibited outcomes, exposure limits, and escalation thresholds.

Assurance

Define the tests, monitoring, incident evidence, and independent challenge needed before and after deployment.

Legitimacy

Include affected people, explain consequential choices, and provide routes for correction and redress.

Questions for a board

A compact governance test

  • Which decisions are being delegated, and which must remain meaningfully human?
  • What can the system access, change, spend, disclose, or promise on our behalf?
  • What evidence demonstrates safe performance outside a controlled evaluation?
  • Who sees weak signals and near misses, and who has authority to intervene?
  • How do suppliers, models, tools, and downstream agents change the chain of accountability?
  • How will affected people contest an outcome or report harm?
  • What would cause us to pause, narrow, or retire the system?

Explore the work

Board and advisory rolesIndependent scrutiny for organisations governing consequential AI. Standards in plain EnglishA practical map of the relevant IEEE standards work. Agentic AI safetyHow governance changes when systems can plan and act. BooksPublished guidance and web galleys connecting technical practice to ethics. ResearchFrameworks, publications, and diagnostic work on machine behaviour. Regulatory challengesAn essay on governing risks that cross ordinary boundaries. The ConstitutionA playable argument about rules, interpretation, and power. Executive briefingsBuild a shared language before a consequential decision.