CIO OPINION deliverables of fairness , accountability , transparency and explainability .
Data governance is all about availability , usability , integrity , and security , whereas AI governance concentrates on aligning AI initiatives with business goals through documentation and audit . These activities are designed to enable accountability , especially as it relates to the potential introduction of bias in ML models .
Both pursue the underlying goal of trust , the trust of the consumer and the trust of the regulator . We have seen the latter ’ s requirements in law for data governance , and in spirit , with laws potentially on the way for AI . In data governance , enterprises must also contend with their own industry standards and any international laws like GDPR .
As we wait for UAE government regulations on AI to take shape , we can reliably speculate on what that shape may be by looking at already published guidelines within the country and the various regulatory and non-regulatory actions of other governments around the world .
Increasingly , businesses must be aware of , and navigate external rules as they digitalise their operations . Understanding where data governance ends and where AI governance begins is vital to remaining on the good side of regulators . When implementing data governance , we look first to policies , centralised catalogues , and stewardship , and strive to maintain data quality .
With AI , things become more complicated , but we can begin with Digital Dubai ’ s ethical guidelines and compose risk-assessment frameworks based on these . We can also review our operational efficiency and produce ways to monitor the value-add of each project . Indeed , the effectiveness of AI governance will hinge on the level of observability of AI systems .
It is in the implementation , where we can see the starkest differences between AI and data governance . AI frameworks should stand apart from cataloguing datasets . AI governance is about enforcing rules around the implementation of solutions to avoid a range of errors during development and deployment .
In digitally competitive economies such as those found in the GCC , a misstep could have severe implications
For example , the US National Institute of Standards and Technology published a range of frameworks , including the AI Risk Management Framework , and the AI Bill of Rights ; the UK launched the AI Safety Institute ; and Singapore ’ s Infocomm Media Development Authority released AI Verify . Even the European Union has an AI Act that empowers the levying of significant penalties for non-compliance .
When implementing data governance , we look first to policies , centralised catalogues , and stewardship , and strive to maintain data quality .
www . intelligentcio . com INTELLIGENTCIO AFRICA 39