“Responsible AI” and “Accountable AI”

As the use of AI proliferates, we are often hearing a demand for “Responsible AI”. We are however repeating that the foundation of AI being responsible is in AI being accountable. This accountability means that there should be an organization that takes up the responsibility to own the consequences of AI. This is precisely what the recent AI advisory for intermediaries from MeitY has done.

It is important to consider that AI cannot be treated as an independent juridical entity. It is either owned by the developer or the company developing the AI algorithm/product or the licensee. The advisory captures this aspect and hence brings in “Accountability” as the leading requirement of AI. Hence “Responsibility” of AI is embedded in the “Accountability” of the licensee/owner.

It is the Accountability that has to be in place to make “Responsibility” an automated consequence.

While on the subject, I came across a discussion on TV about a start up “AI-Kavach” developing a Cyber Security product using AI. The promoter Ms Pratyusha Vemuri was successful in getting a funding support in the Shark Tank India Season 3 and made news.

The discussions on the shark tank is interesting since the company got a valuation of around Rs 20 crores with a funding of Rs 1 crore for a total equity of 5%. The company currently sports a modest 20000 downloads and 1500 paid downloads (at Rs 99 per year subscription) in the consumer segment. I am sure that many know that these downloads occur in a day in most of the cases if the product catches the imagination of the market. But as of today, the downloads in consumer segment seems to have been currently suspended and the Company wants to operate on the B2B segment for the time being. Not sure if this decision was after the MeitY advisory.

We wish goodluck to the entrepreneur for harnessing the power of the AI.

One of the sharks rightly took note that the algorithm has to work in the B2B environment and learn consumer behaviour by profiling which would be used for fraud detection. In other words the business model is to watch the internet and mobile behaviour by sitting on the intermediaries like Airtel and Jio networks of the entire universe of users and derive intelligence from which fraud customers can be identified. This is “Personal Data Mining” and “Big Data Processing”. For Airtel and Jio it would mean selling of personal data of their customers.

While the entrepreneur indicated that a patent is pending in India, I am doubtful whether the novelty feature would be adequate to get the approval of patent.

Using AI based processing to detect and prevent frauds is an established existing Cyber Security activity and therefore lacks patentable novelty. RBI has also made it mandatory through “Adaptive Authentication”. However, the labelling of the product as “AI” has given it a marketable value.

It is not clear if the entrepreneur nor the funding sharks have identified the Privacy/DPDPA risks. However anonymous the process of collection is, such mass profiling carries a high level of Privacy risks.

In another TV interview, the entrepreneur appeared to be banking on anonymisation of the data analysis. I hope the company is able to cover the Privacy Risks with whatever processing they plan to do as part of the AI processing.

The B2B users would however remember that they need to adopt certain compliance measures themselves before adopting such products into their processes.

The Shark tank episode was about a month ago and the AI advisory of 4th March 2024 should be considered as a jolt to the company.

(Comments welcome)

Naavi