There are now credible reports and examples of AI systems writing their own code, in some cases end-to-end, and doing so at a pace that no human can realistically keep up with or fully review.
So let me ask you this…
If systems are evolving faster than we can understand them, and those same systems are being deployed into live environments, who is ultimately accountable?
Directors… It sits squarely with us.
Call me paranoid if you like, but I struggle to see how any organisation can confidently deploy technology that isn’t fully understood, even by the people building it, particularly when it may be generating new logic, structures, and behaviours that weren’t explicitly designed.
Regulation will try to keep up, but realistically it won’t.
This is a leadership challenge.
And at a time where cost savings, speed, and a healthy dose of FOMO are driving adoption, the role of the board is not to accelerate blindly, but to apply judgement, challenge assumptions, and slow things down when needed.
Now is the time to double down on what has always worked. Strong governance, clear accountability, disciplined risk management and proper oversight.
The real risk isn’t AI. It’s overconfidence in deploying something we don’t yet fully understand.
If you’re navigating this at a board or executive level, I’m always open to a friendly conversation about the tools I’m using to tackle this problem.
