Exclusive | Red flags over Kerala’s AI-powered sports push: No tender, no tech experts, and minors' data at stake

The Kerala Sports and Youth Affairs Department has granted administrative sanction for the implementation of the Game Keralam AI & Data Analytics Platform through direct procurement, bypassing the standard open-tendering process. While the overall project outlay is pegged at ₹80 lakh, the state government will directly fund ₹50 lakh. The remaining ₹30 lakh is slated to be mobilised through corporate social responsibility (CSR) initiatives and ecosystem partnerships routed via the Sports Kerala Foundation.

The administrative sanction—issued when Prasanth Nair IAS was serving as special secretary in the department—invokes government orders from 2018 and 2022 to give administrative sanction without an open tender. Under a 2022 Store Purchase Department order, the LDF government revised the financial ceiling for government bodies to directly purchase products and services from Kerala Startup Mission-registered startups without an open tender from ₹20 lakh to ₹50 lakh. By capping the state-funded component at exactly ₹50 lakh, the project has leveraged this provision to award the contract directly to a KSUM-accredited startup.

Conceived as a government-owned digital intelligence layer, the platform is intended to integrate official sports and government datasets to provide role-based dashboards, predictive analytics, AI-assisted recommendations, a natural-language query interface, and secure APIs within an implementation timeline of six to eight months.

However, the decision to bypass a competitive bidding process for a complex data analytics project handling sensitive athletic records—including data involving minors—has drawn scrutiny from industry watchers and governance critics.

Absence of technical oversight

To supervise the rollout, the government has constituted a five-member Project Review Committee chaired by the Sports Secretary, with the Director of Sports and Youth Affairs as convenor, alongside the Additional Secretary, Additional Director, and Finance Officer of the directorate.

Critics, however, point out a glaring omission: the committee comprises solely departmental bureaucrats and finance personnel, with no independent technical expert or data architect appointed to evaluate the platform’s core architecture, assess model validity, or oversee how data will be ingested and utilised.

Operational and legal roadblocks

Technical experts in the analytics domain told THE WEEK that the nature of grassroots athletic data poses major compliance and operational hurdles. They say that state sports and school-level administrative records remain largely low-frequency, fragmented, and confined to analogue paper files or disjointed spreadsheets. Cleaning and standardising these datasets for machine learning models will require substantial engineering bandwidth before any predictive algorithm can function reliably.

They also note that athletic databases invariably involve minors; the platform must navigate stringent compliance requirements under the data protection rules, including verifiable parental consent mechanisms and statutory bans on tracking or behavioural profiling of children. Experts warn that the complete absence of a technical or data-privacy specialist on the government’s Project Review Committee leaves a glaring oversight blind spot, potentially putting sensitive data of young athletes at risk.

Another crucial issue is that tournament and performance data is primarily held by autonomous sports associations and federations that lack standardised APIs and operate under no legal mandate to share proprietary data with the state.

Infrastructure and budget mismatch

The administrative sanction stipulates that the platform must be deployed within Kerala Government State Data Centres (SDCs) using open and interoperable technologies to safeguard data sovereignty, privacy, and vendor independence.

However, experts argue that this requirement creates a critical technical paradox. A state data centre environment cannot easily route workloads through public cloud-based AI APIs (such as OpenAI or Google Cloud) without compromising on-premises data sovereignty mandates. Running localised, open-weight predictive models and natural language query systems on-premise demands dedicated GPU compute and specialized infrastructure—resources seemingly difficult to sustain within the approved financial scope.

According to the departmental estimate, ₹40 lakh is earmarked for solution design, UI/UX, interface development, and system integration, while ₹15 lakh covers infrastructure, DevOps, and security testing. A further ₹15 lakh is allocated toward project management, training, technical support, and first-year operations.

Industry insiders contend that attempting to build, train, deploy, and support an on-premises AI platform within these cost heads points to a significant underestimation of the technical scope, reinforcing questions over why such an ambitious digital initiative was cleared without testing the broader market through a competitive open tender.