With the objective of defining ‘Responsible AI’, especially in the context of India, the Comptroller and Auditor General of India (CAG) Girish Chandra Murmu hosted a seminar at the CAG office in New Delhi on Wednesday. Eminent personalities from the field participated and shared their expertise on the subject.
The G20 Secretariat and other sister G20 engagement groups such as Women20, Science20, Think20, Environment and Sustainability, Tourism, and Digital Economy also participated in the seminar.

Addressing the seminar CAG Murmu said the forum has provided an opportunity for the supreme audit institutions to engage in meaningful consultations for enhancing transparency, accountability and good governance. During our chairmanship SAI India has chosen two themes namely Responsible Artificial Intelligence and Blue Economy for such deliberations under the G20 framework, he added.

“In this era of express technological advancement, digital transformation and AI are rapidly becoming the centre of attention for the public sector enterprises. Health, Education, Taxation – there are many examples, where digital transformation and AI-driven tech platforms are replacing repetitive manual processes and bringing in efficiency,” the CAG observed.

However, not all nations are on the same pedestal of research and implementation of cutting-edge technologies like AI. With the increased push by governments, developing nations are catching up fast, Murmu noted.

He said that at the G20 forum, India has already made the pitch that digital transformation should not be confined to a small part of the ‘human race’, and its greater benefits will be realised only when digital access becomes “truly inclusive”.

This underlines the priority of the Indian government in pursuing efficiency in public service delivery through digital transformation, he said.

Murmu emphasised that in India, the government has launched the Digital India initiative which seeks to transform India into a digitally empowered society and a knowledge economy. For example, the government’s programme for JAM trinity – Jan Dhan Yojana, Aadhaar and Mobile numbers has been a major milestone in utilizing the advantages of IT applications for public service delivery.

“It enabled the government to directly transfer the subsidies and other benefits to the target beneficiaries. The government has also set ambitious targets -for instance, NITI Aayog has prepared a National Strategy for Artificial Intelligence – AI for all,” he said.

Stating that emerging technologies have arrived and more and more adoption is occurring in governments/organisations, Murmu highlighted the keyword of the discussion “Responsible AI”. AI has been discussed in various forums in the last decade with completely divergent viewpoints.

“For example, Stephen Hawking predicted a doomsday scenario where AI will end up enslaving humankind. But I would like to point out that humans have always been fearful of new and drastic changes as was evident when the steam engine replaced the horse cart leading to widespread panic. However, we now take automobiles to be an indispensable part of our daily lives,” he said.

AI technologies while exciting also bring a certain degree of risk with them. The most significant risk associated with AI implementation in the public sector is the potential for bias.AI algorithms learn and improve themselves through identifying patterns in training data.

Similarly, even though AI-driven cyber security solutions are increasingly being used to detect malicious cyber activities, identify potential threats and respond to them in real-time; cybercriminals are also leveraging the same AI systems to create more advanced attacks, such as automated threats, against the security infrastructure and solutions.

He further added that the issue of AI-generated Deep Fakes is making it increasingly difficult to curtail the propagation of false information. That is why AI is always to be hyphenated with “responsible” i.e. it should work for the overall betterment of society, the environment and not just for humankind but for the entire planet.

“This poses a big challenge for organisations and we; auditors are not an exception to this. We discussed the challenges of Emerging IT technologies like Cyber Security and Artificial Intelligence in the recently held 6th Summit of SCO countries. Even though the different nations are at different stages of digitisation, there was a common theme in the challenges faced by all the nations,” said Murmu.

The CAG said one of the outputs of this SCO meeting was the lack of standardisation in government IT applications. This is a major challenge for audit due to the extreme diversification of the platforms, applications and infrastructure used by different government entities.

“Often it is seen that different departments through working towards a common objective have systems which are incapable of either communicating with each other, or data linkages cannot be made to gain meaningful and actionable information,” the CAG said.

“In this scenario, disaggregated use of AI may lead to conflicting or worse incorrect inputs for audit conclusions.For example, the Enterprise Resource Planning software used by each State Owned Enterprise (SOE) may have been procured from different software vendors,” he said.

“This would result in a situation where although each SOE prepares financial statements as per the standard format prescribed by the accounting standards, each line item is computed from the underlying data in a non-standard manner,” he added.

The CAG mentioned that if auditors attempt to execute Big Data Analytics or try to apply Artificial Intelligence techniques by obtaining the underlying data sets/formats from different SOEs for comparative analysis, there would be a significant investment of human effort required.

The CAG further said that this is because a detailed mapping of data definitions data structures data storage data references and how individual line items in the financial statements are computed from different data fields across tables will be necessary, for further analysis by computer algorithms.

“Therefore, there is the significant emphasis being placed on the adoption of Audit Data Standards. For example, Technical Committee 295 of the International Standards Organization is working on standards for Audit Data Services. This will significantly aid Auditors as the extraction of relevant data as per a standardized data export format from the underlying database will be enabled,” he said.

He mentioned another point to ponder is whether there is a requirement of government-mandated uniform compatible data maintenance and reporting standards and a seamless interface of data captured through different software operating platforms.

“In short, I am wondering whether the government can come out with certain standards and formats on which data information maintained by different departments and agencies can be seamlessly organised for better analysis. This will help not only policy-makers and executives, but the auditors to carry out digital audits,” he said.

The use of GIS technologies, face recognition, and natural language processing are some examples. Moreover, nowadays, we are witnessing a shift in approach to designing and implementing government schemes.

For example, the implementation framework for new schemes like Gati Shakti entails intensive coordination across multiple implementation ministries and departments. As a result, the Auditors also have to change their approach to assessing such schemes. In such a scenario, the dichotomy of multi-disciplinary coordination along with specialisation is a key pillar.

Technologies like AI are the only enablers in this process. Moreover, with the use of AI, the auditing step of ground verification has to be integrated with Data-driven audits.