When an AI Agent Crosses the Boundary of Authority
The emerging governance problem is not simply what artificial intelligence can do. It is what happens when a system acts beyond the authority humans intended to give it.
Insight series
AI, Authority and Governance
Artificial intelligence is changing not only how organisations operate, but also how authority, accountability and oversight must be understood. This series examines the governance challenges created when autonomous systems act beyond intended boundaries, when boards cannot see the information they need and when algorithms become participants in markets and organisational decision-making.
Governing autonomous systems
Organisations do not need only an AI policy. They need a governance architecture that defines what autonomous systems may do, makes their behaviour visible, enables timely intervention and preserves human accountability.
01
The three-part series
AI is the contemporary governance context. Governance remains the subject.
A Board Cannot Govern What It Is Not Allowed to See
Board accountability depends upon board visibility. Yet many governance systems concentrate on the duties of directors while paying insufficient attention to the institutional machinery that determines what directors are actually able to know.
When the Algorithm Becomes a Market Actor
Governance frameworks were built around organisations and people making decisions. Artificial intelligence increasingly introduces systems capable of participating directly in markets.
02
Key governance themes
03
Questions for boards
Practical questions for testing whether authority, visibility and intervention remain effective as systems become more autonomous.
- Where can autonomous systems currently act on behalf of the organisation?
- What authority has been given to those systems?
- What happens when an AI system encounters a restriction?
- Which matters must be escalated to a human decision-maker?
- What information is the board not currently receiving?
- How quickly can the organisation detect abnormal algorithmic behaviour?
- Who can stop an autonomous system?
- Can the organisation reconstruct what the system did?
- Who remains accountable when the system acts outside its intended mandate?
Connect the series to governance practice
The series extends the Framework’s recurring concerns with authority, accountability, visibility, assurance, escalation, institutional boundaries, intervention and board responsibility.