`Agentic AI a massive opportunity as well as a potential liability for India’

For India, wherever integer nationalist infrastructure and AI-driven innovation are becoming cardinal to economical growth, agentic AI is simply a monolithic accidental arsenic good arsenic a imaginable liability, said Saugat Sindhu, Global Head, Advisory Services, Cybersecurity & Risk Services, Wipro Limited.

However, helium rapidly added, “Security, privacy, and ethical oversight indispensable germinate arsenic accelerated arsenic the AI itself.’‘

The aboriginal of AI successful India would beryllium defined by the quality of its systems, their spot and the work the state takes to deploy and unafraid them.

According to Mr. Sindhu, agentic AI technologies are reshaping productivity, governance, and nationalist information successful an epoch wherever machines nary longer conscionable assistance but act.

Listing retired immoderate of the astir captious cyber risks of agentic AI, helium said India’s integer system was booming — from UPI payments to Aadhaar-enabled services, from astute manufacturing to AI-powered governance. But arsenic artificial quality evolves from passive ample connection models (LLMs) into autonomous, decision-making agents, the cyber menace scenery is shifting dramatically.

These agentic AI systems tin plan, reason, and enactment independently — interacting with different agents, adapting to changing environments, and making decisions without nonstop quality intervention. ``While this autonomy tin supercharge productivity, it besides opens the doorway to new, high-impact risks that accepted information frameworks aren’t built to handle,’‘ helium cautioned.

There could adjacent beryllium threats involving instrumentality misuse wherever attackers instrumentality AI agents into abusing integrated tools (APIs, outgo gateways, papers processors) via deceptive prompts, starring to hijacking. Or it could beryllium representation poisoning wherever malicious oregon mendacious information is injected into an AI’s short- oregon semipermanent memory, corrupting its discourse and altering decisions., helium explained.

Another signifier of captious menace could beryllium assets overload which involves attempts to overwhelm an AI’s compute, memory, oregon work capableness to degrade show oregon origin failures — particularly successful mission-critical systems similar healthcare oregon transport systems.

Cascading hallucination is different benignant of threat. Here, AI-generated mendacious but plausible accusation made to dispersed done systems, disrupting decisions — from fiscal hazard models to ineligible papers generation.

For example, Mr. Sindhu elaborated, an AI cause successful a banal trading level generated a misleading marketplace report, which was past utilized by different fiscal systems, amplifying the error.