Written by Shubhi Agarwal and Anamika Jaiswal students at Maharashtra National Law University, Mumbai.
India’s gig economy is booming. What began as a handful of ride-hailing and food delivery platforms has rapidly evolved into an ecosystem that, according to NITI Aayog, is expected to employ nearly 23.5 million platform workers by 2029-30. Yet, as this digital workforce expands, the law has struggled to keep pace with the realities of this job.
On 26th April 2025 in Varanasi, when nearly 150 food delivery riders collectively refused to work in temperatures touching 43°C, demanding better pay and basic heat protection such as cotton uniforms. Instead of addressing these concerns, the platform reportedly deactivated their accounts and reinstated them only after they signed written undertakings. The incident exposed the imbalance in bargaining power between platform companies and gig workers and shows how algorithmic management can penalise workers for prioritising their own safety.
The incident also prompted the National Disaster Management Authority (NDMA guidelines) to issue an advisory in July 2025 urging platforms to modify their algorithms during heatwaves, heavy rainfall, and other emergencies by removing penalties linked to acceptance rates and login time, while also providing safety gear and shaded waiting areas. However, the advisory remains non-binding. Since it carries no statutory force, platforms continue to deploy algorithms that reduce incentives, downgrade ratings, or restrict work allocation when workers refuse unsafe assignments. As a result, occupational safety has become completely dependent on corporate discretion rather than it being a legal obligation.
Against this background, this piece firstly examines how AI-driven platforms have transformed the traditional employer-employee relationships; secondly, it compares the emerging legislative responses of states such as Karnataka, Bihar, Jharkhand and Telangana to assess whether existing state laws adequately address algorithmic transparency and accountability and lastly, it argues for a comprehensive central framework that embeds worker welfare, human oversight, and climate-responsive algorithmic governance into India’s rapidly expanding platform economy.
Changing Nature of Employer-Worker Relations
With the continuous developments in AI and despite simultaneous efforts of the government to manage it, its impact on platform workers has been adverse. The discourse surrounding AI has always involved entrenchment into human roles via automation. However, this automation has increasingly led to opaque algorithmic management of workers. Particularly in delivery platforms, AI now acts as a phantom boss managing workers 24/7 by assigning them tasks, setting their routes and performance targets, calculating their pay and deactivating their profile, if needed.
This human-replaced algorithmic boss has neither empathy nor any accountability since the current labour laws in India fail to give gig workers any grievance redressal mechanisms against its decisions. This system of AI-cum-supervisor has, therefore, redefined how we look at the traditional employer-worker relations in an industry.
A detailed study on psychological contract between workers and organisation illustrates this phenomenon of AI anthropomorphising perfectly. The theory of mind articulated here reveals the lived experiences of app-based workers, illustrating how they increasingly perceive and engage with AI systems as de-facto managers. It highlights that workers are in a constant state of negotiation with these systems. However, this interaction remains fundamentally one-sided, lacking the reciprocity that typically characterise traditional contractual relationships.
Moreover, since these systems are efficiency and customer satisfaction based, their approach towards worker safety is lackadaisical. Therefore, weather conditions and risks attached to it are not taken into account by the algorithm, meaning that the system does not slow down or adjust its inputs like penalties on workers, ratings etc during extreme weather conditions. Rather, surge pricing or higher demand can encourage workers to take on more jobs even at the cost of their safety.
This is in line with a study on ego depletion theory which argues how continuous algorithmic control, performance pressure and automated evaluation drains workers’ cognitive resources, reducing their ability to make safe choices under stress. This cognitive depletion may partly explain why nearly nine two-wheeler riders lose their lives on Indian roads every hour. Due to this, the NDMA advisory, which aims to protect workers from extreme weather conditions, falls short as it acts on the assumption that the employer (which is the AI in this case) will make necessary changes in its operating system.
How indian states are dealing with automated systems
Various states like Bihar, Jharkhand, Karnataka, Telangana have legislated on algorithmic transparency. However, state intervention through all these regulations has been limited. Section 13 of the Karnataka Platform Based Gig Workers (Social Security and Welfare) Act, for instance, merely grants workers the right to access information about how algorithms influence their conditions of work, including fares and earnings. However, in reality, this information alone is of limited value if the workers lack the ability to challenge these algorithmic decisions and secure meaningful remedies.
The Bihar Platform Based Gig Workers (Registration, Safety and Welfare) Act, 2025 addresses this gap by granting workers not only the right to obtain information, but also the right to seek a review of algorithmic decisions that affect their livelihoods under Section 13. It further requires platforms to establish grievance redress mechanisms, ensuring timely resolution with human oversight. The Act also provides for designated rest points for workers at key locations within the district if an aggregator or platform engages 100 or more Gig/Platform workers within that district.
Further, a comparison of penalty structures in the above-mentioned legislations with respect to non-transparency of algorithms and non-transparent decision making by platforms reveals the differences in their framework. Jharkhand, taking a behaviour specific, micro level approach, explicitly penalises offences in relation to automated monitoring and decision-making systems, such as failure to disclose and process opacity. Penalty levied ranges from ₹10,000 to ₹2,00,000, with a continuing penalty of ₹10,000 per day (up to a maximum of ₹10 lakh).
Similarly, Telangana’s model also proposes a maximum penalty of ₹5000 for the first contravention and ₹1,00,000 for subsequent contraventions. Although an appreciable step, it lacks in deterring such platforms considering their size and multi-crore turnovers. Such capped penalties may be viewed as a cost of doing business, rather than as incentive for changing the management of the system. Bihar’s formulation in contrast is remarkably weak. It relies on the Welfare Board to determine penalties, without setting any statutory limits or penalties for specific contraventions. This step is fraught with the risk of administrative arbitrariness and legal uncertainty. On the other hand, Karnataka fails to provide any penalties or redressal mechanisms for such opacity at all.
Therefore, all the above-mentioned legislations systematically minimise the gravity of such violations via disproportionate or no penalties, raising concerns regarding regulatory capture and the disguise of protection as provided under these laws. Not only are revenue-based penalty amounts necessary, but also affirmance that algorithmic transparency is not a voluntary, but a legal obligation. Considering current challenges, human oversight also becomes imperative over such decision-making, with algorithms only being in taking minor decisions with respect to workers.
Way forward
We conclude by highlighting a structural challenge in India’s regulation of algorithmic management. Platforms such as Uber, Ola, Swiggy, and Blinkit operate on a pan-India basis through uniform apps and algorithms. However, because labour is a Concurrent List subject, they are increasingly subject to differing disclosure requirements, appeal timelines, language mandates, and deactivation procedures across states. This creates significant compliance costs, legal uncertainty, and operational inefficiencies.
The absence of a comprehensive central law aggravates this problem. Although the Code on Social Security, 2020 recognises gig workers, it remains silent on algorithmic governance, leaving states to fill the regulatory vacuum through piecemeal measures that weaker states may struggle to enforce effectively.
The need of the hour, therefore, is a comprehensive central framework governing algorithmic management across digital labour platforms. Such a framework should require platforms to incorporate factors such as extreme weather, worker safety, and the psychological impact of algorithmic decisions. The employers must move beyond maximising efficiency alone and adopt a multi-factor, worker-centric approach that balances productivity with labour welfare.
Caveat: The views, analyses, and information presented in this article are provided in good faith and for general informational purposes only. No representation or warranty, express or implied, is made regarding the accuracy, adequacy, validity, reliability, or completeness of the information. Readers should conduct their own research and seek professional guidance where appropriate. Neither the author nor the publisher shall be held responsible for any loss, liability, or consequence arising from reliance on this content.
Leave a Reply