When the builders say slow down. The challenge of governing AI at the edge of understanding

If those at the forefront of AI development are questioning whether its progress is moving faster than our ability to understand and control it, how can boards be expected to provide meaningful oversight?

Artificial intelligence is developing at a pace that is reshaping how organisations make decisions, manage risk and define accountability. 

In We Must Pace the Frontier, Dario Amodei argues that AI capability development risks outpacing our ability to understand and control it. This creates an uncomfortable question for organisations deploying AI – how do you govern something that is evolving faster than your understanding of it?

The answer is not to retreat from innovation or create controls so burdensome that they impede responsible experimentation and meaningful change. For boards, the challenge is to encourage a culture that ensures AI adoption remains consistent with the organisation’s risk appetite, purpose, values and stakeholder expectations.

Consider:

Building trust with stakeholders

AI is rapidly reshaping the relationship between organisations and their stakeholders. 

Employees want to understand how AI will affect their roles and opportunities to exercise judgement.

Investors want confidence that AI is being deployed as part of a sustainable strategy – not simply as a productivity opportunity – and that its benefits are being considered alongside risks that could undermine long-term value.

Trust may come less from demonstrating certainty than from establishing how uncertainty will be approached.

Boards should be able to see that management understands where AI is being used, any significant risks, what remains uncertain and how gaps are being addressed.

The role of the company secretary

The Company Secretary can help turn AI from a broad technology discussion into a governance issue the Board can see, question and oversee. This includes ensuring:

The future: Governing the capacity to learn

Boards cannot anticipate or understand every development in AI. They can however govern the organisation’s capacity to learn.

Effective guardrails are not necessarily detailed policies, but capable management, informed directors, a culture that encourages critical thinking and governance processes that can adapt.

The challenge is not to choose between innovation and caution, but to ensure that the organisation’s capacity to govern AI keeps pace with its capacity to adopt it.