Home Technology Security As AI gains autonomy, accountability must go beyond the CISO

As AI gains autonomy, accountability must go beyond the CISO

As AI becomes more autonomous, the boundaries of the CISO’s role are becoming harder to define.

Security leaders remain responsible for protecting technology environments, managing cyber risk, and responding to an increasingly complex threat landscape. Agentic AI adds a different challenge. These systems can pursue objectives, make decisions, interact with other systems, and take actions with limited human intervention. Their value stems not simply from intelligence, but from their ability to act.

As organisations begin deploying these technologies at scale, risk can move at the same pace as the systems creating it.

This is what we mean by agentic velocity. Agentic velocity is not just about speed. As AI systems gain greater autonomy, decisions can be made and acted on faster, leaving organisations less time to assess risk and intervene.

Across conversations with security leaders, technology executives, and business decision-makers in the region, a common concern is how organisations establish the governance, accountability, and resilience needed when autonomous systems begin making decisions and taking action at scale.

A business can delegate authority to an AI agent, but it cannot delegate responsibility for the outcome, nor should that responsibility simply default to security because AI happens to be technology.

Governance has to move upstream

Agentic AI exposes the limits of treating governance as a checkpoint near the end of adoption. Security, risk, and compliance are still too often brought into the process after a technology has been selected or a use case is already taking shape.

That approach becomes difficult to sustain when AI can act autonomously and consequences can unfold at machine speed. Governance therefore needs to operate at agentic velocity. This does not mean matching machines decision for decision. It means establishing the boundaries within which autonomous decisions can be made before systems are given the freedom to act.

In practice, this means defining permissions and access boundaries, determining what an agent can do independently, and setting clear thresholds for when human intervention is required. Monitoring and controls also need to provide visibility into how agents are operating, so organisations can identify when they move outside those boundaries and respond accordingly.

Done well, this should enable innovation rather than constrain it. Bringing business, technology, security, risk, legal, and compliance together earlier gives organisations greater confidence about where autonomy can be used safely. Governance becomes part of how innovation happens, rather than a hurdle it must overcome.

The CISO should coordinate, not inherit

Security leaders understand adversarial threats and technology risk, and are increasingly being asked to translate technical exposure into business consequences. This brings the CISO closer to enterprise resilience and strategic decision-making, without making the role the catch-all owner of AI risk.

No single function can own a technology that cuts across an organisation. Business leaders need to own decisions about where AI is deployed and the value it is expected to create. Technology teams understand how systems are designed and integrated. Legal and compliance teams interpret obligations, risk leaders assess exposure, and data owners determine appropriate access and use.

The CISO can connect these perspectives and challenge assumptions, but ownership should follow the decisions and risks each function controls. Otherwise, organisations risk creating an imbalance where everyone participates in AI, but security becomes accountable when something goes wrong.

Putting these guardrails into practice will require the right technology, but also an operating model where responsibility is explicit. When an AI agent affects a customer, business process, or regulatory obligation, the organisation should already understand who owns the associated risk and how different functions respond when an agent moves outside established boundaries.

For CISOs, the opportunity is not to absorb every new risk created by AI, but to help make those risks visible, translate them into business terms, and ensure the right owners are involved before autonomy is scaled.

Agentic AI will broaden the cybersecurity agenda, but it does not overturn the principles of good governance. Effective oversight, clear accountability, and informed decision-making remain essential. What needs to evolve is how those principles operate when technology can act with greater autonomy and speed.

The organisations best prepared for this shift will establish clear decision rights before AI is given greater freedom to act. Good governance should give businesses the confidence to innovate faster because authority, accountability, and intervention are understood from the outset.

In a world increasingly defined by agentic velocity, the CISO’s greatest contribution may not be owning AI governance. It may be helping the organisation make it operational.

- Advertisement -