The missing human in the AI conversation
Artificial intelligence today sits at the heart of governance, national security, economic activity, healthcare delivery, transportation, finance and public welfare. The global conversation, however, has remained curiously polarised. On one side is unbridled techno-optimism — AI as an inevitable force that must be accelerated at all costs. On the other is a growing anxiety around opacity, loss of control, ethical drift and societal harm. What is missing between these extremes is a practical, operational construct that preserves human sovereignty while still allowing AI to deliver scale, speed and efficiency.
My proposition — first articulated earlier and now more relevant than ever — is simple but foundational: every high-impact AI system must have a clearly identified, trained, empowered and accountable human authority. I call this role the “AI Handler”.
For more defence news, views and updates, visit: Fortress India
This article revisits and expands that idea in the context of India’s evolving national position on AI governance and the global discourse emerging from recent World Economic Forum (WEF) engagements. As India prepares to host the Global AI Summit, this is a moment not merely to articulate principles, but to shape a globally exportable governance architecture — one that is rooted in realism, responsibility and democratic control.
The global context: What Davos is signalling — but not yet solving
Recent World Economic Forum meetings have been unequivocal on one point: AI is no longer a future technology. It is a present-day force reshaping power, productivity and geopolitics. WEF discussions repeatedly emphasised themes such as responsible AI, trust, guardrails, human-centricity and risk-based governance. CEOs, heads of state and regulators broadly agree on the *what*.
What remains unresolved is the how.
Most global frameworks still operate at a principle level — ethics charters, voluntary codes, transparency pledges and corporate self-regulation. These are necessary, but insufficient. They do not answer hard operational questions:
• Who is accountable when an AI system causes harm?
• Who has the authority to override or shut down an AI system in real time?
• Who understands the system well enough to challenge it meaningfully?
• How does a regulator or a court interact with an AI decision chain?
The absence of a clear human locus of responsibility is the central weakness of current AI governance models. Without it, accountability diffuses, litigation explodes and public trust erodes. This is where the AI Handler becomes indispensable.
India’s moment: From principles to practice
India’s recent policy signals mark an important shift. The Government of India’s AI governance approach — anchored in a risk-based, human-centric philosophy — explicitly acknowledges that high-risk AI systems cannot be deployed without meaningful human oversight. This is a critical admission.
India’s digital public infrastructure, scale of citizen-facing platforms, defence modernisation, smart cities, ports, healthcare digitisation and financial inclusion efforts together create one of the world’s most complex AI deployment environments. In such an ecosystem, abstract ethical commitments are not enough. Governance must be operationalised.
The upcoming Global AI Summit positions India not just as a consumer of global norms, but as a shaper of them — particularly for the Global South. India understands scale, diversity, constrained resources and high-stakes public service delivery better than most nations. That experience should inform a new global template.
The AI Handler concept fits squarely into this ambition.
What exactly is an AI Handler?
An AI Handler is not a symbolic human-in-the-loop. Nor is it a ceremonial compliance officer. It is a defined operational role, embedded within the organisation deploying AI, with five core attributes:
1. Authority — the power to intervene, override, pause or shut down an AI system when risk thresholds are breached.
2. Accountability — clear responsibility for decisions taken (or not taken) involving AI outputs.
3. Competence — certified technical, domain and ethical understanding of the AI system under supervision.
4. Visibility — known to regulators, auditors and courts as the human interface to the AI system.
5. Protection — legal and institutional safeguards when acting in good faith under approved protocols.
In essence, the AI Handler restores a basic governance principle that technology disrupted: someone must be answerable.
Why existing models fall short
1. “Human-in-the-loop” is often illusory
In many deployments, human oversight is reduced to rubber-stamping machine recommendations at machine speed. This is not oversight; it is abdication under automation pressure. Without authority, time and institutional backing, humans cannot meaningfully challenge AI outputs.
2. Corporate accountability is uncertain
Assigning responsibility to an organisation alone does not work in practice. Organisations act through people. When no specific individual or role is accountable, failures are explained away as system errors, vendor issues or data problems.
3. Legal systems need a human interface
Courts, regulators and investigative agencies are built to question humans, not neural networks. Without an identified handler, AI systems become legally opaque, increasing litigation and undermining justice.
High-risk domains where AI Handlers are non-negotiable
National security and defence
Autonomous and semi-autonomous systems are already influencing intelligence analysis, logistics, surveillance and decision support. In these environments:
• Speed must coexist with command responsibility
• Rules of engagement must be encoded and supervised
• Accountability chains cannot be ambiguous
An AI Handler in defence settings would operate within classified frameworks, aligned to command hierarchies, ensuring AI augments — not replaces — human military judgement.
Healthcare
From diagnostics to triage and resource allocation, AI decisions can be life-altering. A designated handler ensures that algorithmic recommendations are contextualised, challenged and audited.
Critical infrastructure: ports, power, transport
AI-driven optimisation systems can create cascading failures if unchecked. Human handlers act as circuit breakers, especially in crisis scenarios.
Finance and public welfare
Algorithmic decisions affecting credit, subsidies or fraud detection demand fairness, explainability and redress. Handlers provide a human anchor for citizens and regulators alike.
Addressing the hard questions
Will AI Handlers slow innovation?
No. On the contrary, they enable sustainable innovation. By increasing trust, reducing systemic risk and clarifying liability, handlers make large-scale adoption politically and socially viable.
Who trains the handlers?
India should establish accredited national programmes — public–private in nature — combining technical AI literacy, domain expertise, ethics, law and crisis simulation. Continuous re-certification is essential.
Who bears liability?
A balanced model is required:
• Handlers are accountable for decisions within their authorised scope
• Organisations retain liability for deploying uncertified systems or denying handler authority
• Legal safe harbours protect handlers acting in good faith under approved protocols
Can small organisations afford this?
Not all systems require full-time handlers. Shared, tiered and “handler-as-a-service” models can support SMEs, while high-risk deployments justify dedicated roles.
Why India can lead globally
India is uniquely positioned to shape this idea into a global standard:
• Experience with population-scale digital systems
• A constitutional commitment to democratic accountability
• Credibility across the Global South
• A growing AI ecosystem spanning start-ups to strategic sectors
At a time when global AI governance risks being shaped either by a few technology giants or by fragmented national regulations, India can offer a middle path: innovation with institutional responsibility.
The AI Handler is not an Indian solution for Indian problems — it is a globally relevant governance primitive.
What the Global AI Summit must deliver
If the Summit is to be consequential, it should move beyond declarations and commit to:
1. A clear definition of AI Handler roles across risk tiers
2. National accreditation and certification frameworks
3. Pilot deployments in the public sector and critical infrastructure
4. Legal clarity on authority and liability
5. International dialogue on mutual recognition and incident response
These are tangible outcomes that can reshape global AI governance discourse.
Conclusion: Reclaiming human sovereignty in the age of machines
AI will continue to grow more capable, autonomous and pervasive. The question is not whether machines will assist human decision-making — they already do. The real question is whether humans will remain institutionally sovereign over the decisions that shape lives, security and society.
The AI Handler is a modest but powerful idea. It does not reject technology. It does not romanticise human infallibility. It simply insists that in high-stakes systems, responsibility cannot be automated away.
As India steps onto the global stage to shape the future of AI governance, this idea deserves serious consideration — not as a constraint, but as an enabler of trustworthy, scalable and democratic AI.
If we get this right, India will not just adopt AI responsibly. It will help the world do so.
(Lt Gen M U Nair (retired) is the former National Cyber Security Coordinator, Government of India, and a former Signal Officer in Chief, Indian Army.)
(The opinions expressed in this article are those of the author and do not purport to reflect the opinions or views of THE WEEK)