Regulations such as the Digital Personal Data Protection Act, with an emphasis on explainability and auditability, reflect a growing awareness that trust, and innovation must go parallelly
The headlines today are filled with sci-fi scenarios and speculative doom of artificial intelligence overtaking human control, threatening jobs or even endangering civilisation. But for business and technology leaders, the real threats aren’t decades away. They’re right here in form of fragile, sprawling and hyperconnected systems that are growing more brittle and vulnerable with every wave of AI adoption.
From Imagined Catastrophe to Everyday Crisis
While the public debates revolve around the hypothetical existential threats, organisations are facing an immediate crisis of resilience. The explosion of APIs, multi cloud environments and cloud native applications have created an ecosystem in which data flows so fast that even the most secure measures available today cannot handle it.
For instance, AI tools like ChatGPT, DeepSeek, etc. are often deployed without proper oversight, increasing both capability and risk exposure. And this is an example of every day. AI is now embedded into everything from fraud detection and customer service to app performance, access control, and critical cybersecurity functions. These systems are designed for speed and efficiency but rarely for resilience.
When faced with new cyberattacks, behavioral anomalies or traffic spikes, many AI systems fail. The result is not just downtime, but significant operational and reputational loss.
Fragile Systems Exposing Today’s True AI Risks
According to a report, 96 per cent of organisations are deploying AI models, a fourfold increase since 2023. But only 2 per cent are truly prepared to scale AI securely across their operations. Shockingly, nearly half of all AI proof-of-concept projects never reach production, signalling deep issues in architecture, oversight, and readiness.
Meanwhile, the attack surface continues to grow. In 2024–25, 92 per cent of mitigated application-layer attacks targeted APIs, with many leveraging AI for automated probing and exploitation. Threat actors are corrupting training data, bypassing detection systems, and generating deepfakes that evade legacy defenses. At the same time, unauthorized use of generative AI within organizations is opening doors to data exposure and compliance breach.
India: Scaling Fast, but With Caution
India is advancing rapidly on the AI front. From UPI and Aadhaar to DigiLocker and ONDC, digital infrastructure is enabling billions of interactions daily, many of which are powered or enhanced by AI. Nearly 90 per cent of Indian IT firms are piloting GenAI tools, yet 75 per cent lack structured change management strategies to integrate them effectively.
This is why India’s proactiveness on AI governance is so important. Regulations such as the Digital Personal Data Protection Act, with an emphasis on explainability and auditability, reflect a growing awareness that trust, and innovation must go parallelly.
Resilience Is the Real Competitive Edge
The future of AI depends not only on intelligent models but also on resilient infrastructure. Systems must be secure by design to be able to detect, isolate, and respond to threats in real time, and adapt under pressure. For leaders, it is now critical to bridge gap between AI readiness and operational maturity.
This means start building AI governance into every layer, from development to deployment. AI must operate as an integrated function across analytics, security, and operations, not as isolated experiments. Leaders must also manage model diversity, whether using open-source proprietary or third-party tools, under strict governance frameworks. And as hybrid environments become the norm, infrastructure and security teams must align early to embed resilience at the architecture level.
Similarly, security cannot be an afterthought. While most organizations use AI to strengthen defense, very few have deployed capabilities like AI firewalls or real-time data quality controls. Those that do are gaining faster response times, stronger audit-readiness, and most importantly greater stakeholder trust.
AI Can’t Be Trusted If the Systems It Runs on Aren’t
It is a simple truth: AI is not independent of infrastructure; it depends on it. You can’t build secure, responsible, and scalable AI on systems that are not designed for the speed and complexity of today’s digital environment.
It’s time to shift focus away from speculative AI doom to the real risk infrastructure fragility. Because in the AI era, trust isn’t just a strategic asset it’s a foundational requirement.
-By Pratik Shah, Managing Director – India & SAARC, F5