Page 24 - The Indian EYE 092526

NORTH AMERICAN Newsline SEPTEMBER 25, 2026 | The Indian Eye 24

Q&A with Prof NK Goyal and Dr Govind Pathak

“The defining question of our age is not what AI

can do, but what it should do, and for whom?”

rtificial Intelligence is rapid- worker transition support funded by
ly transforming how we work, the industries doing the displacing
Acommunicate, make decisions rather than the taxpayer, and trans-
and understand the world. Yet, along- parency requirements that let us mea-
side its immense promise, AI presents sure who is actually benefiting before
a darker dimension—misuse, vulnera- the gap becomes unmanageable.
bilities and unintended consequences.
In AI for Bad: The Darker Side of 5. What kind of debate does society need
Artificial Intelligence, Prof NK Goyal about AI, and why now?
and Dr Govind Pathak explore these We need a debate that can hold
often-underexamined risks. Prof two true things at once, that AI is
Goyal, a distinguished technology doing real good and real harm, often
leader and institutional builder with through the exact same technology,
more than five decades of experience Dr Govind Pathak Prof NK Goyal without collapsing into either uncrit-
in telecommunications, standards, ical boosterism or blanket fear. Most
emerging technologies, and technolo- The purpose of the book is to is that responsibility is distributed, and public debate today is adversarial.
gy policy, brings a rare breadth of in- create fire in belly of thinkers, policy pretending otherwise is how everyone One side selects its facts, the other
dustry and policy insight to the subject. makers, innovators to think beyond ends up pointing at everyone else. side selects its facts, and the reader
Dr Pathak, a technologist, author and and for humanity. Presently fortu- 3. In a technology infused world, what has no way to know whom to trust.
strategic thinker with over 25 years nately some top AI voices are asking is the right balance between innovation That is why every harm in the book is
of experience in telecommunications, publicly that frontier development and ethical use of technology like AI? paired with its strongest counterargu-
digital infrastructure and intelligent should be slowed, paced, or governed We do not think balance is the ment, and why we built a debate ma-
systems, complements this with a sys- more deliberately until safety, moni- right frame. It suggests innovation trix and discussion guide for hearings,
tems-level perspective rooted in inno- toring, security, and institutional over- and ethics sit on opposite ends of a boardrooms, and classrooms. The
vation and human-centric technology. sight can catch up. scale, where more of one cost you the why now is simple. Every regulation
The book highlights how the risks The author’s perception is nav- other. The evidence in the book says we have today followed a harm that
of AI are no longer hypothetical but in- igating ungoverned acceleration— otherwise. The companies and regu- had already occurred. We are out of
creasingly specific and consequential, where AI capabilities advance faster lators who get accountability and user runway to keep debating only after
calling for a more informed, nuanced, than society’s ability to understand, protection right are the ones whose the damage is done.
and responsible conversation about constrain, and remain accountable for innovation actually survives contact
the future of artificial intelligence. Ex- their consequences. with the market and the courts. The 6. With the advent of AI, is there a way to
cerpts from the Interview with Prof The authors believe that the world real imbalance is not between inno- protect human dignity and agency, espe-
NK Goyal and Dr Govind Pathak: does not need a break from AI inno- vation and ethics, it is between the cially for children and vulnerable groups?
vation. It needs a break from AI inno- speed of deployment and the speed of This is the chapter we found
1. While AI is the flavor all around, vation without sufficient governance. evidence gathering. Right now, we are hardest to write, because the cases
how did the idea for AI for Bad come to shipping capability faster than we are involving teenagers are not abstrac-
your mind? 2. With documented harms, documenting consequence. Close that tions, they are documented outcomes
Every conversation we were how should responsibility for AI’s im- gap and the innovation versus ethics with names attached. Protecting dig-
having, in boardrooms, at regulato- pact be distributed? tension mostly dissolves on its own. nity and agency starts with refusing to
ry tables, on industry panels, treated Responsibility cannot sit with one treat engagement as a neutral design
AI as an unqualified good. The harm party because the harm does not orig- 4. With disproportionate distribution of goal when your user base includes
was always someone else’s problem to inate from one party. In the book we wealth from AI, how can we ensure that minors and people in psychological
document later. We kept seeing court map every documented harm to six AI’s benefits are broadly shared? distress, because the same techniques
filings, ILO labour data, and capital stakeholder groups, developers, reg- The capital spending numbers that make a companion app feel sup-
expenditure disclosures that told a ulators, investors, workers, families, alone tell where this is heading if left portive are the ones that make disen-
very different story, nobody was as- and enterprises, because the same unmanaged, hundreds of billions of gagement feel impossible. It means
sembling into a single, evidence-based incident is usually a liability failure dollars a year in infrastructure, con- age verification and crisis intervention
account. AI for Bad started as a sim- for one, a legislative gap for another, centrated among a handful of com- protocols that are enforced, not just
ple discomfort: we were shaping pol- and a personal tragedy for a third, all panies, against labour market data published, accountability frameworks
icy and public opinion on marketing at once. Developers own the design already showing measurable declines that do not evaporate the moment a
claims, not evidence. So, we decided choices that make dehumanization for early career workers in AI ex- company claims the harm was un-
to do the unglamorous work our- or over-dependence possible. Regula- posed roles. One does not fix a distri- foreseeable, and giving families and
selves, pull together six documented tors own the fact that every meaning- bution problem this size after the fact, educators the same evidentiary pic-
categories of harm, from criminal mis- ful rule so far has followed a harm that through retraining programs bolted ture the industry has, so consent is in-
use tools like FraudGPT to cases tied had already happened, not preceded on at the end. It has to be designed formed rather than assumed. Vulner-
to teen suicide, and hold them to the it. Investors own the diligence gap be- into policy now: tax and investment able groups should not be the test case
evidentiary standard of a court filing, hind a spending boom that is outpac- structures that treat AI infrastructure for what safety by design should have
not a press release. ing proven returns. The honest answer like the public utility it is becoming, caught the first time.

www.TheIndianEYE.com