India has emerged as the world’s second-largest country for AI adoption, just behind the United States, according to a report by cloud computing security firm Zscaler, which warned of escalating security risks and threats from so-called “artificial intelligence proxy” (AI).
And the report stated that India ranked second globally in AI and machine learning transactions within institutions, after analyzing nearly one trillion AI and machine learning transactions during 2025 through the Zscaler Zero Trust Exchange, during the period from January to December 2025.
And India recorded 82.3 billion transactions between June and December 2025, accounting for 46.2% of total activity in the Asia-Pacific region, topping the region. The highest levels of accreditation are concentrated in the technology and communications, manufacturing, services and financial sectors.
And Zscaler’s report comes just weeks before the India AI Impact Summit 2026, scheduled to take place in Bharat Mandapam, New Delhi from February 16-20, 2026, with the participation of prominent global tech leaders, including NVIDIA CEO Jensen Huang, OpenAI CEO Sam Altman, Google CEO Sundar Pichai, Anthropic CEO Dario Amodi, and Qualcomm CEO Cristiano Amon.
And despite the strong momentum in AI adoption, the report stressed that India faces significant security challenges, including arming proxy AI and using it in advanced cyber attacks.
And the report pointed out that the growth of artificial intelligence in India is driven by central government policies supporting digital transformation, investments in artificial intelligence infrastructure, and skills development, as well as the expansion of qualified manpower and cloud computing companies capable of rapidly deploying artificial intelligence services, which contributed to strong growth compared to previous years.
And the largest activity of artificial intelligence was concentrated in the technology and communications sector (31.3 billion transactions), followed by manufacturing (15.7 billion), then services (12.6 billion), and the financial and insurance sector (12.2 billion). But he cautioned that many organizations do not even have a baseline inventory of their active AI models, putting their sensitive data at serious risk.
Suvaprata Sinha, resident information security adviser (CISO-in-Residence) at Zscaler in India, said the pace of adoption of AI within the country’s institutions outstrips the ability of most organizations to govern it.
“As AI is integrated into everyday business applications and workflows, the security priority for Indian organizations is clear: understand where AI is being used, examine the data in circulation, and consistently enforce appropriate controls. A zero-trust approach, with robust data protection and continuous visibility, is essential to secure the AI-driven transformation as fast as the market demands it.”
And according to the report, offensive tests showed that enterprise AI systems fail quickly, with critical gaps emerging within minutes. And the average time of the first critical failure was 16 minutes, while 90% of the systems were breached in less than 90 minutes, and in some cases the means of defense were exceeded within seconds.
Data Leak Risks
And the report highlighted that data leakage remains one of the most prominent concerns, as more than 18,000 terabytes of data were passed to AI applications globally in 2025, equivalent to about 3.6 billion digital images. Grammarly recorded about 3,615 terabytes of data traffic, while ChatGPT with 2,021 terabytes turned into one of the largest stores of sensitive corporate information in the world.
And ChatGPT alone has been linked to some 410 million data leak prevention (DLP) policy violations, including attempts to share sensitive information such as code and medical records.
Debbin Desai, Executive Vice President of Cybersecurity at Zscaler, said: “AI is no longer just a tool for raising productivity, it has become a major vector for self-attacks and rapid machine-level attacks, both by cybercrime and by states. And in the era of proxy AI, an attack can go from detection to lateral movement and then data theft within minutes, rendering traditional defenses useless. To win this race, organizations must confront AI with AI, deploying a smart zero-trust structure that closes all potential paths to attackers.”
comments