Learn, explore and unleash your inner chef.

“AI is evolving so quickly that most people don’t fully comprehend how transformative it can be. Organisations need to ensure they do not misuse or mismanage AI, that the right internal procedures and frameworks are in place to truly benefit from AI instead of introducing new risks,” says Henry Umney, Managing Director, GRC Strategy, Mitratech.

Mitratech is hosting the marcus evans Online Event: Harness the Opportunities Offered by AI while Ensuring Effective Controls and Governance.

What is possible with AI today?

There are a whole host of opportunities e.g. content creation, automation, tailoring of products to customers, etc, but we still don’t fully understand the extent of what is possible. At the moment, we are just scratching the surface of what capabilities AI and machine learning can offer.   

When deploying AI, what should they be careful of?

There are several things. We have heard of ChatGPT hallucinations. The best way I would describe AI is to consider it like a new assistant. In the old days, you would ask your new hire to put together a report, but you would never just present it to your manager without checking it first. The danger is in thinking that computers are infallible. There is also the risk of bias, incorrect or incomplete input data, not calibrating the model correctly, which can easily give the wrong answer. There are many pitfalls. We need to be very careful of the data being used by the model, and work out a way to test inputs and outputs, to ensure they’re working within the agreed parameters.

In what ways are some companies misusing or mismanaging AI? 

There are some very well-known cases. For example, if you want to use a model to scan CVs to narrow down the list of suitable candidates and you use your current workforce as the basis of a “good hire”, the outcome will be to propagate any historical bias. As such, you need to take care in setting the parameters to get the outcome you want. Data loss and data privacy are also a key concern and risk. For example, engineers at a technology company uploaded a number of technical specs of their new product to ChatGPT and asked it to write a user manual. Whilst this was successful, all of that confidential information is now in the public domain. As more and more employees start using AI, data privacy, confidentiality and bias are just a few of the risks you need to be wary of.

How can organisations feel more empowered?

Visibility. Does the organisation know and understand all the AI models being used internally, whether its their own or that of a third party? Secondly, does the organisation understand the risks those models pose? Has it ranked the risks? Does it know what models the business relies on? Thirdly, it is about making sure it has controls in place to mitigate risks and work within the guidelines. That presupposes that there is a governance framework in place. Mitratech has a solution that helps people embed an AI governance framework in the organisation, create an inventory of applications, risk rank, and make sure the right controls are in place based on that framework.

Whilst technology can be a huge enabler, organisations need to ensure the right governance framework is in place defining roles and responsibilities ideally with a board member accountable for the programme.

Any final thoughts?

AI is evolving so quickly that there’s almost a hesitancy to start the governance journey. A pragmatic first step is to do an initial assessment, understand what AI is being used, and the maturity of any controls already in place. We anticipate multiple standards to be released by regulators, countries, trade bodies, in different industries and jurisdictions in the next couple of years, giving global companies a complex web of reporting to navigate, so having a technology enabled framework will make this more easily achievable.