In the AI era, cybersecurity and data resilience strategies are undergoing a structural reset. Traditional security frameworks, built for predictable systems and human-led interactions, are increasingly being challenged by AI-driven environments where autonomous agents, unstructured data, and machine-generated decisions are becoming central to enterprise operations.
The larger concern now revolves around trust, whether the data feeding AI systems is reliable, whether autonomous systems can be governed responsibly, and whether enterprises can maintain resilience in increasingly fragmented, multi-cloud environments.
In an interaction with Express Computer, Sandeep Bhambure, MD and VP – India and SAARC, Veeam, discusses how AI is fundamentally altering cyber resilience priorities, why enterprises are struggling with fragmented trust architectures, and how governance, observability, and trusted data ecosystems are becoming critical in the AI era.
AI is forcing enterprises to rethink traditional security models
According to Bhambure, the rise of AI and autonomous systems is disrupting long-established security assumptions across enterprises. “If you look at the agentic era, the entire security landscape and definitions are kind of just being thrown out of the window,” Bhambure says.
He explains that many of the trust and security frameworks built during the pre-AI era are no longer sufficient for modern AI-driven environments. “Whatever trust and security infrastructures that were built in the pre-AI era can no longer be useful in the post-AI era,” he notes.
Bhambure points out that enterprises are now being forced to revisit not only their AI adoption strategies, but also the underlying governance and trust mechanisms supporting them.
The real challenge lies in unstructured data and AI trust
Bhambure believes one of the biggest reasons AI projects continue to struggle is the quality and trustworthiness of enterprise data itself.
“A large number of AI projects are failing because the data which is feeding these AI projects cannot be trusted,” he observes.
He pointed out that enterprises are no longer operating in the relatively structured world of traditional business intelligence systems. AI systems now depend heavily on unstructured data, documents, conversations, images, videos, emails, and other contextual datasets that are growing rapidly across enterprises.
“The BI era was all about transactional data, structured data, but AI is all about unstructured data,” Bhambure avers.
The scale of this challenge is massive. He notes that global data creation is accelerating rapidly, with the majority of enterprise information now expected to remain unstructured.
The explosion of unstructured information is also creating new attack surfaces, especially as AI agents themselves begin interacting autonomously with enterprise systems. “You are no longer protecting against humans, but you are also protecting against AI agents,” Bhambure warns.
Autonomous entities are already beginning to outnumber human users significantly inside enterprise environments. “In an enterprise, autonomous agents and entities outnumber humans by a ratio of 82:1,” he points out.
Compliance burdens are increasing in the post-AI era
As regulations such as India’s Digital Personal Data Protection Act gain momentum, Bhambure believes enterprises are facing a far more complex compliance environment than before. “Organisations have the responsibility of complying with DPDP in a post-AI era, and that is making the challenge even bigger,” he says.
He points out that enterprises can no longer treat compliance, governance, privacy, and resilience as separate operational silos. Instead, they need a far more integrated and comprehensive approach to data governance.
Many organisations currently operate fragmented security and governance architectures, using multiple disconnected tools for privacy, governance, security, consent management, and compliance.
As a result, enterprises are often creating additional operational complexity while attempting to solve governance problems independently across different systems.
“They are trying to put in fragmented bits and pieces of solutions, which is creating another nightmare to manage,” Bhambure adds.
Data classification and observability are becoming foundational
Bhambure believes observability and contextual understanding of enterprise data are becoming central to resilience strategies.
He asserts, “Enterprises first need to understand which datasets are genuinely mission-critical, which datasets involve personally identifiable information (PII), and how different datasets map to regulatory obligations.”
Enterprises also require contextual intelligence around how data is accessed, who is accessing it, and whether those users are humans or autonomous agents.
This contextual visibility becomes even more important because enterprises are increasingly concerned about AI-generated errors, hallucinations, and data poisoning incidents. “The biggest fear for people rolling out AI in a big way is the mistakes that AI can make,” he says. Enterprises therefore need the ability to reverse or recover from AI-driven mistakes quickly and precisely.
Enterprises are still stuck in fragmented AI trust architectures
Bhambure believes the industry still lacks a unified approach towards AI trust, governance, and resilience.
As a result, enterprises continue approaching AI governance through fragmented product stacks and disconnected frameworks.
This fragmented approach often increases operational complexity rather than simplifying governance.
Cyber resilience talent development is becoming equally important
Beyond enterprise technology adoption, Bhambure also highlights the widening cyber resilience skills gap in India.
Industry demand for cyber resilience talent is growing much faster than the number of trained professionals entering the workforce. To address this, Bhambure says multiple collaborations are underway with universities and educational institutions to improve awareness and hands-on exposure around cyber resilience and data governance concepts.
“Our goal is to create a pool of a hundred thousand skilled resources,” he points out.
He also stresses the importance of practical learning environments and early industry exposure for students entering cybersecurity and AI-related fields.
“We are creating 100-plus centres of excellence across colleges and universities where students get to dabble with technology,” Bhambure adds.
AI trust and resilience will dominate enterprise priorities
Looking ahead, Bhambure believes enterprise technology priorities will increasingly revolve around enabling trusted AI adoption at scale.
However, he emphasises that successful AI adoption will depend heavily on building robust trust and governance layers around enterprise data ecosystems.
“It is our responsibility to ensure that organisations see the ROI and success of unleashing the power of AI,” Bhambure notes. For this, enterprises must move beyond prolonged experimentation cycles and focus instead on smaller, outcome-driven AI deployments that demonstrate measurable value.