Jobs at stake: Legislation could shield employees facing an AI threat

This Artificial Intelligence (Protection of Rights of Employees) Bill is narrower in scope than its European counterpart, the EU Artificial Intelligence Act, which intends to cover AI risks in general. The Indian private member’s bill, in contrast, focuses squarely on employee rights in the workplace.

Also Read: The AI revolution could leave women even further behind. Here's how.

To understand the need for such legislation, look no further than US tech giants. Amazon, for example, has been developing machine learning algorithms since 2014 for warehouse logistics, pricing decisions, etc. One of these works as a recruitment tool by scoring candidates on a scale of one to five stars and then picking the top resume from the hundreds received for every Amazon job.

In 2015, the company discovered that its algorithm favoured male candidates, as it had been trained on successful resumes of the preceding 10 years, which were mostly of male applicants. The algorithm had ‘learnt’ that male candidates were preferred. Resumes of women with pursuits such as ‘chess’ or ‘debate’ were given an even more unfavourable rating. Although Amazon initially tried to amend its algorithm, it ultimately abandoned the project.

The Indian bill attempts to tackle this problem by proposing the “mitigation of biases in AI algorithms that may adversely affect employees’ rights or opportunities by regular conduct of performance audits and impact assessment."

Also Read: Outrage over AI is pointless if we’re clueless about AI models

Additionally, it attempts to replicate the EU’s General Data Protection Regulation (GDPR), whose Article 22 provides that “the data subject shall have the right not to be subject to a decision based solely on automated processing, including profiling, which produces legal effects concerning him or her or similarly significantly affects him or her." In this context, the Indian bill’s text requires ‘consent’ to be obtained from an employee whose data is to be utilized.

While it is a step in the right direction, the bill in its current form faces challenges.

Audit and overview: The Indian bill mandates organizations to conduct an Equality Impact Assessment at five-year intervals to ensure “fairness, non-discrimination, and compliance with established rules and regulations." A five-year period in the context of AI development is like an eternity. The EU’s AI Act requires the registration of any ‘high risk’ AI system to increase transparency. Such timely registration in India would be a good idea.

Also Read: AI didn’t take the job. It changed what the job is.

Consent: India’s Digital Personal Data Protection Act (DPDPA) emphasizes consent as the primary enabler of personal data processing. The Trinamool member’s AI bill too mandates ‘explicit’ consent in writing from employees who would be affected by an AI algorithm. But achieving explicit consent in a country as economically and educationally diverse as India would pose employers a challenge. While the DPDPA suggests the use of ‘consent managers’ to ensure consensual flow of personal data between data subjects and fiduciaries, how employers could ensure explicit and continued consent remains unclear.

Employer versus employee rights: The DPDPA grants employers the right to process employee data for ‘legitimate uses’: “for the purposes of employment or those related to safeguarding the employer from loss or liability, such as prevention of corporate espionage, maintenance of confidentiality of trade secrets, intellectual property, classified information or provision of any service or benefit sought by a Data Principal who is an employee."

Conversely, the AI bill under discussion requires Indian authorities to protect the rights of employees where they “refuse tasks or decisions solely based on AI-generated processes if they feel these violate their rights or ethical standards." It also grants them the right to review high-risk decisions made by AI processes, affording them an opportunity to “challenge choices or decisions adversely impacting their employment."

A possible clash is envisaged between the rights of an employer that wants to use employee data and the rights of employees who want such data removed.

Also Read: Devina Mehra: Yes, AI is coming for our jobs; it’s the old story of new tech

AI implementation for payroll reduction: Perhaps to address concerns around impending layoffs arising from the adoption of AI-based models by Indian businesses, the proposed legislation requires the provision of “adequate training and up-skilling opportunities for employees affected by AI implementation" to ensure that “they can effectively adapt and engage with the technology." Given that AI models are often inducted by employers to reduce their payroll costs, it would be counter-intuitive to require organizations to incur expenditure on those it may intend to retrench. While well-intended, this clause would result in additional business expenditure.

The AI bill proposed by Mausam Noor, while more narrowly focused than its EU counterpart, attempts to provide a crucial legal foundation for safeguarding employee rights in the face of increasing AI integration in the workplace and aims to address concerns about job security, discrimination and privacy. It also underscores the need for transparent, fair and human-centric AI development and deployment in a relatively labour-intensive economy such as ours.

The author is an independent privacy lawyer.