Myntra, Flipkart and others build AI firewalls as retail data risks rise

Artificial intelligence adoption is the bedrock for growth in India’s apparel retail sector today. After two years of aggressive investments in generative search, AI styling assistants, predictive inventory systems and automated customer support, online fashion platforms are now focusing on a more fundamental challenge: controlling how AI accesses and processes sensitive business and consumer data.

India’s digital retail, currently estimated at $65-66 billion in gross merchandise value (GMV), is projected to reach $170-180 billion by 2030. However, as AI becomes deeply embedded into marketplace operations, companies are shifting technology spending from rapid experimentation to enterprise-level data governance.

The introduction of stricter enforcement mechanisms under India’s Digital Personal Data Protection (DPDP) Act, which allows penalties of up to Rs 250 crore for major security failures, has increased this shift. For fashion platforms, where customer preferences, body measurements, browsing behaviour and purchase histories power recommendation engines, data protection has become a strategic business priority.

AI expansion meets data risk

Fashion marketplaces rely heavily on consumer intelligence. Unlike many other retail categories, online apparel depends on highly personalised recommendations because purchase decisions are influenced by fit, style preferences, seasonality and previous browsing patterns.

However, the same data that improves conversion rates also creates significant compliance risks. Customer information today extends beyond names and delivery addresses to include sizing profiles, style preferences, payment identifiers and behavioural data. At the enterprise level, fashion platforms also manage sensitive commercial information such as vendor agreements, supplier margins, inventory strategies and upcoming promotional campaigns.

The challenge emerges when large language models (LLMs) are integrated directly with enterprise databases. Without strict controls, AI systems can develop what technology teams describe as ‘permission inflation’, accessing information beyond the authority level of the employee using the tool.

For retailers, the consequences can be significant. Apparel contributes nearly half of online fashion revenues, but the sector continues to struggle with return rates of 25-35 per cent, largely due to sizing issues. AI-powered fit recommendation engines are therefore commercially valuable, but training these systems on unrestricted consumer body data can create regulatory vulnerabilities.

From AI access to zero-trust issues

Fashion platforms are responding by redesigning their AI infrastructure around zero-trust principles. Instead of allowing AI models unrestricted access to central data repositories, companies are implementing Role-Based Access Control (RBAC), where artificial intelligence tools inherit the same permissions as the employees using them.

Under this model, a merchandising executive analysing category performance would receive only the information relevant to that role, while customer service teams would see transaction-level details without access to confidential supplier or financial data. This approach is becoming critical as platforms integrate AI across multiple functions from demand forecasting and inventory optimisation to seller management and customer support.

Table:  AI integration balancing innovation with security

Myntra’s privacy-first AI model

Among India’s leading fashion platforms, Myntra has emerged as an example of how large-scale AI adoption can be combined with enterprise safeguards. The company has developed BIRA (Business Intelligence Retrieval Agent), an internal AI retrieval system designed to help employees access business insights without providing unrestricted access to underlying databases.

Unlike traditional AI deployments where models directly interact with enterprise information repositories, BIRA operates through an intermediary layer that evaluates metadata, employee permissions and organisational hierarchy before processing requests. For example, a merchandising manager analysing festive collection performance may receive access to inventory and sales insights, while a customer support executive querying the same system would not be able to view supplier pricing or confidential business information. As per Pramod Patil, Chief Technology Officer, they have built a platform where privacy is not an afterthought, the company’s approach is embedding privacy controls into the architecture itself. The company has also deployed AI applications such as its customer support assistant Meera, which handles over 30 per cent of standard queries, while supply chain simulations that previously required two days can now be completed within an hour.

Personalisation without privacy trade-offs

The commercial case for AI in fashion remains strong. Online apparel retailers operate in a highly competitive scenario where personalisation directly influences conversion. Natural-language shopping assistants, AI styling tools and recommendation engines are helping platforms improve discovery and increase basket sizes. Consumer interactions with conversational shopping tools have shown higher purchase intent, while AI-generated outfit recommendations can encourage greater cross-category purchases.

However, retailers are moving towards privacy-preserving personalisation. Instead of feeding raw customer information into AI systems, companies are creating anonymised preference profiles, tokenised consumer signals and controlled data layers. This allows platforms to continue improving recommendations without exposing personally identifiable information (PII) to external AI models. “Retailers can no longer evaluate AI purely on engagement uplift or conversion speed; data governance is now an active component of balance-sheet protection,” says Anirudh Sen, Senior Retail Technology Consultant at RedSeer Strategy Consultants.

Protecting the marketplace

The data challenge becomes more complex for multi-brand platforms such as Myntra, Ajio, Nykaa, Flipkart and Meesho, which operate large seller ecosystems. Thousands of brands connect their inventory systems, pricing tools and logistics platforms with marketplace APIs. Without strict controls, external AI tools could potentially extract competitive pricing insights, customer behaviour patterns or marketplace intelligence. To address this, platforms are deploying isolated API environments, restricted data-sharing agreements and zero-retention policies for external AI services.

AI models used for fraud detection and return optimisation are also being limited to transactional indicators rather than broad consumer profiles, allowing companies to reduce losses while protecting customer privacy.

Data governance a competitive advantage

As Indian online retail expands deeper into Tier-II, III markets, AI will remain central to improving discovery, logistics and customer experience. However, the next phase of fashion e-commerce growth will depend less on who deploys AI fastest and more on who deploys it responsibly.

For fashion platforms, privacy-first AI is no longer simply a compliance requirement. It is becoming a foundation for consumer trust, operational efficiency and long-term enterprise value. Companies that successfully balance personalised shopping experiences with disciplined data governance will be better positioned to lead India’s next decade of digital fashion retail.