AI governance is about balancing innovation with protection of privacy, says Wipro’s Ivana Bartoletti

The very nature of AI, particularly generative AI, challenges some core concepts like privacy. GenAI is built by scraping the web, which has a lot of information, including personal information. It challenges the essence of privacy since it’s built on a huge amount of data while any privacy law will tell say the key principle for privacy is minimised data.

GenAI is a big challenge to how we’ve seen privacy, which is also about agency, dignity, and individual autonomy. When algorithms make decisions about individuals, they may lead to two things — discrimination and the loss of human agency.

The concern around AI is it challenges the concept of privacy as a human right, something that has been championed in India. Privacy is a human right. While AI is capable of great stuff, it is also capable of stuff that is conducive to harm to individuals, especially women. Think about deepfakes used to denigrate women.

The intersection between AI and privacy needs to be worked on and studied. This is why we are working on responsible AI and AI governance.

We started early on because AI never existed in isolation. We set up an AI governance task force two years ago to bring the company together on an AI governance journey that is within our governance construct. AI governance is a corporate responsibility.

The main challenge we wanted to tackle was vocabulary, language, and skills. A problem with AI governance is people coming from different backgrounds — technical, legal, procurement, risk management, speak different vocabularies. The first thing was to get everybody on the same page. For example, lawyers should be able to understand coding and technical people, the law, because we must embed the law into the machines.

The entire task force was trained in an AI governance program under the International Association for Privacy Professionals so everybody in that group could understand each other. We trained over 200,000 people on the basics of AI and on responsible AI immediately. They were told AI is not neutral and that algorithmic-driven discrimination can lock people out of services. We also embedded the governance structure across the whole organisation. Whether you’re purchasing or developing an AI product, you must follow due diligence — assessing the risks of AI.

If something is a high risk, we wrap a lot of mitigation around it and try to support the teams, whether internally, or client-facing, to ensure they build the protection into the projects and the systems. But the challenge is to embed privacy, security, legal protection, and equity into the systems at the design stage. This is where we put a lot of effort.

There is an assumption the data about you available is accurate. If I Google somebody, I would imagine the data that shows up is accurate. And if it’s not accurate, it might need to be altered.

There is a conflict between accuracy and AI because it hallucinates. As a citizen, you want accurate information about yourself. We must bear in mind that these models may hallucinate. While it is improving, it’s not completely gone. This may impact the accuracy of the outputs.

In my view, hallucination is the wrong word because it’s too human. Human beings can hallucinate, not machines. There is a separate language for humans and machines. Describing machines like you would a human doesn’t teach people that both kinds of intelligence are different.

These errors are not hallucinations, but statistical mistakes. In other instances, words like AI companionship can be referred to as machine-enabled support systems instead. A big challenge today is to realise we want to grow with AI and have machine and human collaboration. To do that, we can’t deceive people. It is important to omit an anthropomorphisation of machines since they’re different from us.

Following the Paris AI Summit, the next global AI summit will be in India, which is a big thing because it is a testament to India’s ambition of playing a global role in the AI landscape in terms of infrastructure, data centres, skills, and education.

India AI aims to create a robust ecosystem supported by government policies that promote innovation, growth, economic growth, and inclusivity. Prime Minister Modi also announced the IndiaAI mission, and an investment in research, and open-source AI at the summit. He said India wants to play a role in using AI for innovation.

India can play a crucial role here because, for example, bringing together innovation and responsibility by creating AI that does not infringe upon our ethical norms, breach human rights, or discriminate. It means creating and investing in long-term AI that people trust and use. And the more people trust AI because it’s done properly, the more countries grow because their citizens are using it.