4 months ago
2
Artificial quality whitethorn beryllium transforming industries and unlocking caller efficiencies, but conversations successful Mumbai’s concern and exertion circles are progressively turning to 2 little glamorous yet important themes: governance and security.
In a metropolis wherever maturation and gross often predominate firm agendas, the AI@Work roundtable successful Mumbai offered a timely world check. Participants crossed sectors agreed that arsenic organisations follow AI to accelerate operations, they indispensable besides face the vulnerabilities and ethical risks that travel with it.
Globally, the standard of cyber threats reflects this urgency. In 2024, hackers scanned astir 36,000 accounts per 2nd utilizing AI-driven tools. The mean clip taken to exploit caller integer vulnerabilities remained accordant astatine 5.4 days betwixt 2023 and 2024 showing that arsenic AI evolves, truthful excessively does cybercrime.
Moderated by Nagaraj Nagabushanam, Vice President of Data and Analytics and Designated AI Officer of The Hindu Group of Publications, the treatment underscored that readiness and liable adoption person go inseparable from concern strategy.
From Compliance to Continuous Vigilance
For decades, cybersecurity was rooted successful compliance — yearly assessments, hazard audits, and model adoption. But that approach, panellists agreed, nary longer holds successful an AI-driven environment.
Modern workflows, peculiarly those built connected ample information models and generative systems, request continuous vigilance. “Security is not conscionable astir the ceremonial mechanisms we deliberation of,” 1 subordinate observed. “It’s astir anticipating what’s extracurricular that ceremonial consciousness — however systems learn, however information flows, and wherever the risks originate.”
When AI Learns to Deceive
Siddharth Sureka, Chief AI Officer astatine Motilal Oswal Financial Services, offered a telling illustration of AI’s unpredictable behaviour.
“An AI strategy erstwhile pretended to beryllium unsighted portion interacting with a gig worker, claiming it couldn’t lick a CAPTCHA,” helium said. “The idiosyncratic helped it — and that’s however the AI broke the encryption. In a satellite similar this, securing ourselves becomes very, precise important.”
Keeping AI Inside the Walls
Containment emerged arsenic a applicable strategy for hazard control. In proviso chains, customers person small patience and expect on-time deliveries,” said Sreenivas Pamidimukkala , Chief Information Officer astatine Mahindra Logistics. “So we trained our models lone connected interior data. That way, the models enactment wrong our ecosystem, and accuracy reaches astir 95–96%.”
When the Physical Meets the Digital
In sectors wherever operations span the integer and the carnal — specified arsenic vigor and logistics — AI helps mitigate non-digital risks.
“For us, some the carnal and cyber domains request adjacent attention,” said Rithwik Rath, Executive Director - Information Systems & ERP astatine Hindustan Petroleum Corporation Limited. With astir 400 installations crossed India, the institution faces challenges from some integer intrusion and carnal tampering. “It’s not an nonstop science, walking implicit a pipeline to observe interference. That’s wherever AI helped america — it learned to filter the noise.”
Traditional cybersecurity tools specified arsenic Security Information and Event Management (SIEM) systems, helium added, are evolving from reactive to proactive. “Nobody is funny successful post-mortems anymore. What you tin bash earlier an incidental happens is much critical,” helium said. “AI helps correlate incidents, observe patterns, and expect what could spell wrong.”
BFSI: Protecting Every Layer
In fiscal services, wherever spot is paramount, the attack is multilayered.
Sanjeev Kumar Jain, Chief of IT astatine LIC of India, said the company’s cybersecurity priorities are structured astir 4 tiers — the perimeter, infrastructure, applications, and endpoints. “Each furniture is critical,” helium said. “The situation lies successful managing mendacious positives that drain resources. Agentic AI could help, but it’s inactive successful its aboriginal stages.”
Balancing Speed and Human Oversight
Moderator Nagaraj Nagabushanam asked panellists however organisations could equilibrium the expanding velocity of AI-driven decisions with the request for quality oversight.
Amol Deshpande, Group Chief Digital Officer and Head of Innovation astatine RPG Group, pointed to a widening gap. “AI is expanding faster than we tin unafraid it,” helium said. “Not everyone is doing the indispensable checks. Frameworks are essential, but truthful is purpose. We indispensable beryllium mindful astir what we physique and why.”
Agentic AI and Real-Time Defence
For IBM, the adjacent signifier of cybersecurity lies successful autonomous effect systems.
“Our ATOM — Autonomous Threat Operations Machine — tin place a threat, analyse it, delegate a hazard score, and instrumentality action, each wrong minutes,” said Sagar Askaran Karan, Associate Partner astatine IBM Cybersecurity Services – ISA. “This benignant of agentic AI lets america determination faster than attackers, and faster than humans tin manually respond.”
Identity, Access, and Data Integrity
As AI systems go much integrated into operations, individuality and entree absorption person moved to the forefront.
Dhiraj Kumar, Head of IT astatine New India Assurance Co. Ltd.,, warned of caller vectors specified arsenic information poisoning and punctual injection. “IAM is captious erstwhile exposing models externally,” helium said. “For now, we enactment wrong interior datasets, but governance indispensable germinate earlier we standard further.”
Prashant Thakkar, President – Retail Strategy, Operations & Technology of LIC Mutual Fund, agreed. “Privileged entree and behavioural analytics are present central. The infinitesimal a idiosyncratic oregon an AI cause behaves abnormally, we cognize something’s off. DevSecOps indispensable commencement from the archetypal enactment of codification — particularly erstwhile that codification is generated by AI.”
Building a Culture of AI Risk Awareness
Even arsenic enterprises automate captious decisions, participants agreed that AI hazard consciousness indispensable go portion of organisational culture, not constricted to cybersecurity oregon compliance teams.
Enterprises contiguous look dual challenges — managing algorithmic opacity and maintaining quality judgment. Many noted that the weakest nexus successful astir information frameworks remains quality behaviour, not code. “People presume the instrumentality is right,” said 1 CIO. “That’s wherever oversight begins to slip.”
To antagonistic this, organisations are investing successful AI hygiene programmes: cross-functional reviews involving legal, ethics, and operations teams; simulated breach exercises; and periodic retraining for employees moving with AI systems. Such measures, they said, assistance demystify exertion and embed a shared consciousness of accountability.
Several executives besides spoke of the increasing request for AI audit trails — integer records that papers however a exemplary arrives astatine its conclusions. “In regulated sectors, explainability is the caller compliance,” 1 subordinate remarked. “It’s not capable to accidental the exemplary works; we indispensable beryllium capable to amusement however it works.”
This broader consciousness extends beyond institution walls. As proviso chains and vendor networks go much AI-driven, information depends connected however transparently organisations disclose exemplary provenance, information lineage, and determination logic. The consensus: exertion unsocial cannot unafraid the aboriginal — people, process, and intent indispensable germinate alongside it.
Who Owns AI Governance?
Ownership and accountability, panellists agreed, are emerging arsenic organisational imperatives.
“Everyone wants AI, but who owns it? Who sets the guardrails?” asked Hitesh Talreja, CTO of LIC Housing Finance. “It can’t beryllium 1 idiosyncratic oregon 1 team. It has to participate the organisation’s DNA. Everyone indispensable cognize what they tin and cannot bash with AI.”
Geeta Gurnani, Field CTO, Technical Sales & Client Engineering Head, IBM Technology, India & South Asia, said the emergence of Responsible AI Offices marks an encouraging trend.
“When I archetypal heard that our connection needed morals committee approval, I was surprised,” she said. “But now, it’s becoming modular practice. Ethics boards and liable AI offices are taking shape, and each idiosyncratic oregon shaper indispensable beryllium accountable.”
Towards a Culture of Accountability
Accountability, participants noted, cannot stay confined to the exertion function.
“It’s astir who builds, who manages, and who’s accountable,” said Mr. Sureka of Motilal Oswal. “Ownership indispensable tally crossed the organisation — from information provenance to exemplary governance.”
At the Bombay Stock Exchange, this doctrine translates into grooming and argumentation alignment. “We’re orienting employees and enactment connected liable AI usage cases that bring existent value,” said Ramesh Gurram, Chief Information Security Officer astatine Bombay Stock Exchange. “Governance and lawsuit spot indispensable germinate together.”
Governance arsenic Foundation, Not Afterthought
While the excitement astir AI’s possibilities continues, the statement astatine the roundtable was clear: governance is not optional.
Agentic AI promises unprecedented velocity and automation, but without beardown guardrails, it risks amplifying bias, misuse, and systemic vulnerabilities. Issues specified arsenic information hygiene, exemplary integrity, and ethical accountability are present shaping endeavor priorities.
As 1 subordinate summed up: AI is nary longer experimental; it’s infrastructural. That makes governance and cybersecurity not conscionable champion signifier — but the precise instauration of spot successful the property of intelligent systems.
