Can We Trust AI Blindly?

Last week, as the world flocked to DeepSeek given all the justified hype around it, many in India pointed out that the Chinese chatbot gave thorny issues a wide berth.

Ivor Soans

11 Feb 2025, 04:56 PM IST i

<div class="paragraphs"><p>AI cannot be trusted blindly today, but it’s imperative we enable and build systems with trust at the core if AI has to reach its true transformational potential. (Image source: Freepik)</p></div>

AI cannot be trusted blindly today, but it’s imperative we enable and build systems with trust at the core if AI has to reach its true transformational potential. (Image source: Freepik)

Dear Subscriber,

Welcome to Beyond Tomorrow!

We are in week 11 of our new weekly newsletter, crafted to keep you at the forefront of AI advancement. Every week, we’ll bring you a round-up of the most impactful AI stories, highlight how they shape our world, and provide insights into what’s next in the ever-evolving AI landscape.

If you’d like to receive Beyond Tomorrow in your inbox every week, click here to subscribe.

Last week, as the world flocked to DeepSeek given all the justified hype around it, many in India pointed out that the Chinese chatbot gave thorny issues a wide berth. For instance, DeepSeek was reluctant to discuss what happened on June 4, 1989, at the Tiananmen Square and went mum when it came to other 'sensitive' queries on China. DeepSeek also seems to lack safety guardrails and users could prompt the chatbot to generate malicious output. There are some serious questions over security too.

The DeepSeek AI Revolution Has A Security Problem

How DeepSeek Can Be Weaponised: From Bomb-Making Instructions To Revealing Malware And Malicious Code

One would have thought that seeing DeepSeek pussyfoot around 'sensitive' topics would have opened our eyes to the fact that AI tools can be unreliable—AI is often only as good as the data it is trained on and the manner it has been trained.

But humans are often inadvertently overawed by advancements in technology. This may result in some of us trusting AI chatbots blindly. This hit home for me during an argument on a WhatsApp group. A well-meaning friend decided to ask an AI chatbot what it thought, and the chatbot spewed out an answer. And he presented it as if it were empirical evidence. I might have accepted it except he also presented the sources. And when I looked into it, I realised they were surprisingly poor. One was a study that extrapolated data from a survey of less than 2000 Indians to over 750 million other Indians. And the survey did not pass the smell test either as I soon realised it wasn't conducted by an independent agency but rather commissioned by an industry player and hence was likely designed to align with their business priorities. Another source was equally suspect. And the third was not exactly relevant to the topic on hand but the AI tool had included it because some factors seemed to intersect. But no harm done.

However, as AI is increasingly integrated into mission-critical systems and infrastructure and takes decisions, basing a decision on bad data or limited training could prove disastrous. It is crucially important to build effective guardrails and put checks and balances in place. Some of this may need to be legislated in our best interest. Leave alone mission-critical systems, we need to reduce the risk of AI taking a wrong decision on a loan application or making an erroneous suggestion on a personalised treatment plan due to poor quality data.

AI cannot be trusted blindly today, but it’s imperative we enable and build systems with trust at the core if AI has to reach its true transformational potential. That’s one key factor that will empower IT decision makers to trust their systems to take the right business decisions too.

Data Integration Woes Hurting AI Adoption; 80% Red Flag This As Major Challenge

Here are some of the other must-read insights and news from this week when it comes to the world of AI:

OpenAI's Sam Altman Wants India To Develop End-To-End AI Solutions

Ola Founder Bhavish Aggarwal To Invest Rs 2,000 Crore In Krutrim, Launches AI Lab

New Acrobat AI Assistant Gen AI Features Banish Contract Confusion; Huge Benefits For Consumers, Businesses

OpenAI Releases AI Agent Designed To Act Like A Research Analyst

To Make India Self-Reliant For AI Needs, Yotta Launches DeepSeek-Powered myShakti Sovereign Gen AI Chatbot

ChatGPT-WhatsApp: Image Upload, Voice Message Support, Replies, And More Now Available Via Updates

Elon Musk Faces Long Court Battle Over Falling Out With OpenAI

Tata Steel Vouches On Integration Of AI And Industry Technologies To Enhance Output, Quality

NBFC Poonawalla Fincorp Leaps Into AI-Driven Audit And Governance With ServiceNow For Tech-First Fin Ecosystem

ITC Launches Global Centre Of Excellence For Artificial Intelligence Near Kolkata