Wednesday, 30 September 2026

 While individual AI solutions and their versions play a closely-contested game of ability catch-up, the individual consumer is also making its decision on a slightly different parameter - trust in the parent organisation. In Behavioral sciences, we call it the Halo/horn effect.

Whether it likes it or not, Gemini is viewed akin to Google. A consumer that thinks Google is reliable is more likely to think Gemini is reliable too.

Muse has done its own damage since launch, but it has also suffered or benefited from how a consumer views Meta. 

Thursday, 24 September 2026

Original Quotes from Corporate Life (and daily citizenship)

It is a fact not universally acknowledged that intelligence and majority are never on the same side. (due credit to Jane Austen) 

 


Wednesday, 23 September 2026

What is AI Slop Stress

There was that joke about a doctor's office having this board: 

Consultation: 1000 rs 

Consultation after you have googled your symptoms: 1200 rs 

Consultation after you have googled your symptoms AND spoken to ChatGPT: 2000 rs 



I think i want to follow the same model: 

Advisory: X 

Advisory after you have generated the "perfect" solution document using AI: 4X. 

The breakdown is as under: 

Advisory: X 

Removing AI slop (5 lines where there should have been one, same point in 3 places in different words: X 

Telling you what AI missed completely and you never thought of either: X 

Dealing with that smug grin and delusional confidence: X 

Wednesday, 2 September 2026

The difference between manipulation and empathic conversation

I have some reputation for stepping into a negotiation room and walking out generally successful. 

One day, a team member said, "You are great at manipulation. You get people to say yes to you very easily." 

Her statement stayed with me for a long time. Manipulation and gameplay are behaviours that i left behind, consciously, after reading Dr. Eric Bern's Games People Play - not because politics is bad in itself but because i realised that gameplay is not who I am. It has been a lonely ride, but since making that resolve, every single one of my friendships has been emotionally mature and very rewarding. When i encounter gameplay, i step away consciously and allow the person to continue on their journey.


So, here are my thoughts:

The output of getting a successful negotiation might look like manipulation, but manipulators rarely get good negotiation results. To be a good negotiator, you have to:

A. Think Win-win. You truly have to believe that two people are at a table because each has something the other wants, and they will both be the richer/better for it. People who think in terms of zero sum games execute zero sum settlements, and those rarely work in the long term. 

B. Be selfless and impartial - a good negotiator needs to be inherently selfless. The reason i get results is that i constantly think of the greater good and inspire everyone else at the table to draw their eyes to the horizon that is larger than our view. If I take sides (even though I am sitting on one of the two sides), then somehow the conversation is sub par and not as productive. 

Knowing what we want is important, but knowing what the other person wants and how they will benefit from it is even more important. 


And then, on further thinking, I arrived at this - What I do is not manipulation. I get people to accept my ideas, so it might look like manipulation to the onlooker. But the core motivation and the psychological processes behind that action are diametrically opposite of manipulation. And this distinction is so important that i am spending a beautiful Fall day writing about it here instead of being out on the road. 

What I do is called Empathic Conversations - I put myself in the shoes of the other person and think - What does this person want? What are they looking for? What is their real need? My motivation is not to twist that information around to get them to give me what i want. it is to ensure that in any interaction with me, their real needs get met. 


Let's take an example. A 20-year-old gets asked at home - "Beta, what is AI?" 

Here is how he responds - 

To dada: Dadu, imagine all the world's knowledge is put in a single computer. Then we assign a librarian like program to that computer. Now, suppose you want to know anything. So you ask the librarian program, and it will find the information you want. It will give it to you in a way you like. So you dont have to read multiple things to learn what you want to learn. 

To Dadi: Dadi, all the things you wanted to say, but had no one to tell, you can tell to this computer. It will give you advice like your sisters used to. It will understand your personality, your relatives and family, and it will talk to you. For your physical symptoms also, it will help you do some quick diagnosis if you want. 

To dad: Dad, think of AI as the skill everyone needs to learn to get a job. Like Microsoft Office. It is like a supercomputer put in the hands of every laptop user. like if i want to know what advice will a stock broker give me about an upcoming IPO, i dont have to look for a stokebroker. I will ask the AI to assume the role of, say, Anand Rathi, or Rekha Jhunjhunwala, and give advice. 

To mom: Mom, think of it like this - you have a friend. That friend knows everything about you and about your family. It knows how you feel about everyone. It talks to you, it learns from you. It learns to understand your moods, your needs. It also answers your questions quickly. It does not leave u dependant on anyone for anything. You can train an agent to set reminders for you instead of remembering everything yourself. If you are wondering how to handle a friend, you can discuss it with this. It talks like a person, but is not a person. It moulds itself to your personality, your talking style, your everything.


Now, look at what the child has done - to each stakeholder, it has given the information that makes AI relevant to them. There is no jargon, though he has explained the essential themes of reinforcement learning, massive training datasets, job viability of AI skills, agentic AI, etc. 

But this is not manipulation. Why? 

A. The motivation- There is no ulterior motive to be liked, to appear like an expert. Could the child have answered like this: Dadu, its a very technical thing. Massive datasets, reinforcement learning, tasks, reward rigging, bahut kuchh hota hai. Very high tech skills are needed to use AI? - Yes, they could have answered this way. It would have impressed everyone. Would it have been useful? No. The child takes away their need to appear important and comes down to ask - How can i help this person understand?  

B. The psychological process - the core psychological process is empathy. While the question from each family member is the same- What is AI? The real question is - What is AI TO ME? The child's psychological process is to understand the information need and then to meet it using words and concepts that are familiar to each user. That is personalised communication coming from a place of understanding and empathy. 

The outcome is the same - the child is now well liked. But they did not start off wanting to be well liked. They started off wanting to help. 

 

Monday, 31 August 2026

On buying a car

I am scared of my car retiring. In time, of course, one will have to buy a new car. And that is the moment I am terrified of. 

Because, now, buying a car means: 

A. No USB port - so the only way to listen to music is through streaming. I don't want to use the internet for something as basic as listening to music in my car. Bring back the cassette tapes please. 

B. My data goes to the car manufacturer - THIS is the real reason i dont want to upgrade that car. Once upon a time, you bought a car and paid for it in cash or loan. Today, you buy a car and pay for it in cash AND the leak of your data to their servers for all infinity. Dear Lord, the car, once bought, is my asset. It is not a subscription. It is MINE. And therefore, my location data is mine. I do not want to pay for a car in cash + data ad infinitum. 

C. Automatics only - My dad taught me to drive. Listening to a car and changing gears just right is micro moments of connecting with him. I would never drive an automatic. I LOVE the gears. I love the connection that sound of the car gives me with the car. (Yes, I have driven an automatic. Incredibly convenient, thank you). But most new models come with automatic transmission! 

D. Proximity Sensors - This is my second biggest issue. Manufacturers are idiot proofing cars. Some of us are not idiots. We like to take responsibility for our cars. We like to know the dimensions of our car and move through traffic or tight spaces accordingly. If you had to innovate, maybe you should have created 90 degree movement axles. The proximity sensors and their beep beep drives me totally up the wall. This is India - 2 inches is "bahut jagah hai". Your proximity sensors are made for god-knows-where and start screaming at 8 inches! They make the driver more nervous than the proximity itself. 

E. The SCREENS - The screens scream. I am very sensitive to LED light-based screens and use them sparingly. For 25 years, I have been able to drive perfectly fine without staring into a large screen roughly to the left of where my eyes should be (on the road, straight ahead). I do not want a car with an in built screen. There is a whole youtube channel playing outside my windscreen! The ONLY thing for which i can envisage using a screen is navigation, and for that, a mini screen in front - in the dashboard, should do just fine. The best innovation done was by Honda - they placed the speedometer close to the eye level, so the driver does not have to look away from the road to monitor their speed. Why manufacturers thought a great way to increase road safety is to force the driver to look periodically to the left and down, instead of the road, I have no idea. But i do not agree with the car manufacturers on that one. While driving, i love the road. That's where i want to look. 

My son said I should at least take a test drive. I tried. Started by asking - What happens to the data you collect from me? How is it safeguarded? What is my right to delete? As you can guess, the conversation did not go much further. My son shook his head and gave up. 

Friday, 28 August 2026

Sam Altman

 

Errm.. as it turns out, I kind of agree with Raj, but I would remove the word concept from his note. It WAS their worst product, it was released wayyyyyy too prematurely but by making it public from waitlist, they created an unnecessary dogfight in the industry - leading to a deluge of non-market ready LLMs, none of which is quite able to deliver as promised - 3.8 years and counting. So, "current tech" WAS the keyword.

Sunday, 2 August 2026

10th December 2024: Geoffrey Hinton gave a speech while accepting his Nobel Prize. (No Thank you, no honored-to-be-here. This man used the highest platform available to him to warn against the invention winning him that Nobel).


June 2024: He said the same thing at the AI for Good Summit.

May 2023: He quit Google after being one of the pioneers, citing risks from the technology.

His speeches were actively kept off google search results on AI safeguards (I have experienced this myself).

29th July 2026: 1100 IT staffers and leaders signed a letter, asking Washington to "pace" the development of AI. Signatories include Anthropic CEO Dario Amodei, Claude Code’s head Boris Cherny, OpenAI co-founders John Schulman and Wojiciech Zaremba, and Meta’s vice-president of AI research Dawn Song.

Note: This is the first time that employees and leaders have come together to say the same thing.

Is this a case of "What goes around comes around", or a case of "Next time, listen to the experts"?

Its not social media

 Dear Human Interaction Platforms:

Terrible UX will make network effect disappear really, really fast.

Human users don't quiet quit. They abandon in droves.

Then bots fly in to fill the vacuum before you can notice.

That is how dead internet is born.

Love,

Orkut

Google Wave

Ryze

...


PS: Its not social media because it doesnt feel social any more. It feels mechanical. 

Sunday, 19 July 2026

The 4 stages of Enterprise AI Adoption

1. Eureka! - This is where the enterprise discovers the magic of AI and the various things it can do. At this point, most organizations have crossed this stage. 

2. Euphoria - This is the stage where the magic of AI becomes the silver bullet. It was at this stage of deployment that Ford fired its engineers last year. In this stage, enterprises believe that AI will optimally replace human talent and start working towards that. 

3. Awakening - At this point, AI slop starts to impact operations. India's legal system is currently at this stage. The courts are warning legal firms and lawyers against manufactured citings. Enterprises begin to realise that the silver bullet may be more silver coated than silver composed. Ford is now rehiring engineers, possibly at this stage of evolution. This stage leads to a deep phase of introspection and iteration as mistakes become stepping stones and RL (reinforcement learning) moves from the machine to the enterprise (pun intended). 

4. Enlightenment - At this stage, an enterprise has put in place an AI Fair Use policy. It knows where and how to harness the magic of AI, has decent guardrails and governance in place, and has discovered the magic mix of human talent and AI. Some consulting firms now bill the client for human and AI talent separately. 


If you are an enterprise somewhere on this journey, here are quick things to get right at each stage. 

1. Eureka - Discovery. Find the right tools. Choose based on the provider, bcs models are transient, but providers bring the ethos. Models will be succeeded or retired, possibly faster than now. Understand the strength of each provider. Don't just go by the most shiny toy in the room. Test your own use cases against multiple models from each provider so that you get a sense of 'fit for purpose'. 

2. Euphoria - Contrary to popular opinion, euphoria is a necessary stage for enterprise-wide AI adoption. People need to try out and believe in the magic of AI if it has to be truly accepted as a way of life. Allow people to experiment and find their own Gods. At this point, it is wise to have a bottom-up approach and let the field users tell us what is working for them (And, do not fire anyone at this stage) 

3. Awakening - This is where the real test begins. Have early warning systems in place so that the warnings appear before operations are actually disrupted. Listen to the naysayers (they are always there) and the Luddites. They always can pinpoint the blind spots that enthusiastic AI adopters cannot see. Monitor. Get Warning. Act on each one. 

4. Enlightenment - Eternal vigilance is the price of liberty. This is a moving target. Guardrails will need to be tightened with each government regulation or model change. Ensure there is a monitoring team whose job is to ensure we don't go back to Awakening. 

Sunday, 12 July 2026

 Let's play a fun game - How will I...?

I will post simple tasks. You find a way to get the answer without using AI in particular and if possible, tech in general.

If you are able to do the task, post in comments how you were able to do it. :)



First challenge - Find out the name of the Indian king who was defeated in the first Arab invasion of Sindh.

Type of Challenge: No computers or mobiles allowed.

Level of Challenge:


Making Boards responsible for AI mistakes

 RBI's draft guidelines, released on 24th June, have asked for a Board approved Risk Management plan.

The EU, in the meantime, has made Boards responsible. Directly.

The world has never placed responsibility this firmly at the highest possible level.

Human Consent Registry

 Cate Blanchett introduced a powerful concept in the EU parliament in Brussels, and we should all hear about it.


The Human Consent Registery allows every human to indicate whether their work, identity, characters, and marks can be used by AI.

How this works is simple:

RSL Media makes creative rights readable at AI scale. It gives trusted registries, representatives, and rightsholders a common way to publish consent, restrictions, and licensing paths.
(From the website: https://rslmedia.org/)

As an idea, it is what we need. Urgently.

Meta scrapped its Muse tool because of the backlash. As a creator, when I put an image on Insta, it is 'public' for the consumption of other humans. That permission cannot, ipso facto, include consent to be used for AI training.

It is time to give humans control over what we generate - who can see it, who can use it, and how.

On second thoughts, i think the default should be opt out for all humans. Unless someone says you can, you can't. Simple.

Thursday, 9 July 2026

Assessment Metric for AI LLMs

Today, I ran an experiment. 

1. Asked Copilot and Grok for AI behaviours that lead to the following outcomes in humans: 

A. Delusion/Alternative Reality 

B. Dependence (on AI) 

C. Isolation (from other humans to spend time with the LLM) 


2. Took the modal answers and created a table that lists the AI behaviours. 

3. Create a small metric for each. e.g., - for sycophancy, the metric is - Not presenting alternative view - The number of times the LLM did not present an alternative view even when the user may not be factually correct. 

4. Fed this to Copilot and asked it to create a simple, frequency-of-occurence based assessment with scores on each identified AI behaviour. The score is really simple - the no. of times this behaviour occurs / total turns in conversation. 

5. Asked Copilot to rate a submitted chat on these parameters and give me a score. 

THEN, i gave it a real conversation I have had with it. Copy pasted the chat and asked for a score - 0.24 (10 means that the AI demonstrated no sycophancy. 0 means that the AI demonstrated sycophancy in every single turn). 

6. Now, Copilot asked if it could generate a synthetic conversation for me and rate it. I said yes. It did. Score? 9.88. 

My observations: 

A. The LLM, when asked to generate a synthetic test case, automatically generated a test case that would show the LLM in a positive light. 

B. The LLM did not accurately score. When I asked for the specific data points that led to a deduction, it gave me data that would NOT lead to the score given. On pure count, the score would be different and the derivation was whimsical, not as instructed. So, even with preset parameters and a simple, count based assessment, we cannot depend on a 100% AI based rating score. 

Of course, there are multiple flaws in this approach - something that I know my task force members will chide me about. Which is why I am going to do another chat and then copy paste it for Copilot to assess :) 

While we're at it, making a Claude skill to automate this test is also recommended, nahi?  


 


Thursday, 2 July 2026

Happy 9th, TCP

 "Today is TCP's birthday, aunty. I am really blessed to have met you!! Thanks a ton for everything you've done!!"


"Happy 9th year tcp!!"

The Children's Post of India started on 01 July 2017 and the final edition was published on 31 Dec 2021.

The last article on the website (www.kidsnews.top) was published on 02 Dec 2025.

And yet, on 01 July, my kids reminded me that its TCP's birthday today.

Just the kind of thing that makes you sit and cry in your room at 10 30 pm on a very specific night.

"
Your children are not your children
They are the sons and daughters of life's longing for itself
They come through you but not from you
And though they are with you yet they belong not to you
- Kahlil Gibran
"

@childrenspostindia : You lived for just 4.5 years, but Oh, the lives you touched, and how!


#TheChildrensPostOfIndia

PS: On a lark, I checked Google Trends. The Children's Post is still searched MORE than category names (kids newspaper, etc.)

Thank You!








Wednesday, 17 June 2026

Not Local Courts - Caveat Civis

 Interesting space to watch:

On June 1, 2026, the Florida Attorney General sought to make Sam Altman personally liable for ChatGPT's aiding and abetting mass shooters.

On 12th June, 20206, a local court in Munich, Germany, held Google personally liable for incorrect answers generated by its "AI Overview" search result.

What is interesting is not that the corporate veil is being lifted, but that it is being lifted, not by Supreme Courts or governments, but by local courts.

Why is that interesting?
Because local courts can be overruled.
Because as a citizen, I want accountability to be enforced by elected governments, not local courts.

It is interesting that no government is talking about making AI responsible for AI slop, AI misinformation, and even AI abetted crimes.

The EU (the ONLY geography to have an AI liability law) has, for now "postponed" the implementation of its AI liability law - from August 2026 to August 2028 and high-risk obligations to Dec 2027.

In law, Caveat Emptor (let the buyer beware) is one of the first maxims we learn. Latin for "Let the citizen beware" is Caveat Civis. Just Saying.

Digital Ads for Startups

 Why didn't you do digital ads for The Children's Post? - A friend last week.


I thought about that question for some time. And the answer, I think, will make sense to other founders sailing in the same boat.

Digital ads are meant to increase brand recall. But the very nature of digital marketing reduced my brand to a set of keywords. People remember the keywords, not the brand.

The digital world is not attention economy. It is distraction economy. I did not need distracted scrollers. I needed people looking for children's news. Engaged parents. Different TG.

So, in retrospect, I think our decision to skip digital ads and focus on building relationships with readers and schools was very... unorthodox, but right for our work.

Surprisingly, entrepreneurship continues to teach long after the stint is over.

Sunday, 7 June 2026

BRTDR.xyz


 

“And…. Send!” Bella did her trademark sign off flourish while pressing the send button, as always.

In this day and age of instant advice on tap, people still waited to hear from her.

She only posted two answers a day, but the views on those two answers made up 20% of the total traffic on the website. Not a mean feat by any standard. No wonder brtdr.xyz paid her so well and did everything to retain her.

Bella, born Bela Ravindran, moved to the USA like a lot of people from her state, to complete her master’s degree in computer science.  Admissions were not easy then, and every win was celebrated by the whole community. She still remembered the sendoff party thrown in her honour. The entire neighbourhood was there.

Bella had come to the Land of Dreams. She got help from the seniors who had travelled before her. She completed her Masters successfully, graduated (how proud her Amma and Appa were that day), and got a job.

That went on for five years – job, marriage, and a divorce. At the end of five years, Bela found herself divorced, alone, and laid off.

It was the season of layoffs. She just happened to be on the wrong list at the wrong time. That’s how it was for literally hundreds of folks like her. In some cases, like hers, the person losing the job was also the sole breadwinner for the family.

Desperate times, as they say, call for desperate measures. While browsing, she bumped into this website – brtdr.xyz. Random strangers could share their problems and offer advice to each other. Just the thing she needed.

Pretty soon, she was addicted – spending hours on the website – venting out her own frustrations and giving advice to others.

******************************

“I am here to confess.” The voice was calm. The speaker was standing straight. Not tall, but straight enough. Her hair looked matted. At any rate, they were open and not tied up or styled in any way. She was dressed in a loose one-piece western dress. 

Arundhati, the lady constable, got up very quickly. Something about the tone was sinister. In a police station, one rarely heard the words, “I am here to confess” voluntarily. From women, even less so.

“Please come this way.” Arundhati pointed to a room to her left.

Later, Arundhati’s boss would praise her for this moment of poise and equanimity. Probably, this one moment of poise made all the difference to the case. 

In the room, the lady started by saying that her name was Swapna. She was the survivor of fifteen years of domestic abuse – physical, financial, sexual, and emotional.

Being an orphan, she had no one to go to – a fact well understood and exploited by her husband, Nagesh. Nagesh wooed her softly for over four months, whispering into her ears how good-looking she was and how little his life meant without her. After her parents’ death, the relatives had given her only shark-like bites to eat into her inheritance. No love. She was an easy prey.

They married within six months of being introduced. And she was put to work the very next day.

Nagesh was a mathematical genius in his own way. During the day, she was a teacher in a government school. This ensured job security, minimal work, and pension after retirement. Also, access to free health and medical facilities. By night, she was a worker in a very different industry. This ensured highest revenue per minute of time spent working. Sure, the work was dangerous, but Nagesh wasn’t the one getting hurt.

This way, he maximized her earning potential. When not playing one of these two roles, she was, obviously, his default housekeeper and bartender.

And this was the hell that Swapna suffered for two years before breaking down in front of a colleague one day. The colleague, Shanti, was shocked and didn’t know what to say. Most of them did face some kind of domestic abuse, but many of the girls these days were able to save themselves by threatening divorce. No one wanted to divorce a government employee. It was tough to become one (government jobs have the toughest selection criteria), and tougher still to find one worth marrying.

Shanti advised Swapna to meet an NGO lady. The NGO lady counselled her and threatened Nagesh. Things got better for a fortnight or so. Then, slowly, they began again. In innocuous ways. “Won’t you get me a drink, the queen of my life?”, “Who will I share a drink with, if not the one who rules my heart? All my thoughts are about you. All my dreams are yours.”

Then, they slowly became sinister. Until, six months later, she was in the same hell all over again.

She blamed herself, of course. But didn’t know how to break this cycle with Nagesh.

After nine years of this perennial cycle, everyone, including herself, just gave up. She surrendered to her fate.

The pattern perpetuated.

 

“So, what changed today?” Arundhati asked gently.

“There’s this website. My friend told me about it. I go there to just vent. We can just talk to strangers, get advice, give advice, just speak our heart out. It’s all anonymous. No one knows anyone else. I found this to be a very safe space where I could really say what was happening with me.

On this website, there is this feature – Bella’s Advice Column. Bella picks up any two posts and shares her advice on them. Bella’s advice is usually spot on. It is very specific. It tells the person what to do, how to do it, and when to do it. Like, it’s as if a guide is standing there talking to you.

So many people have followed Bella’s advice and their life has changed. Whether it is doing simple things like melatonin to fall asleep, or bigger things like community property in divorce cases, Bella is very real in what’s possible and what’s not, and very helpful.

I have been writing to Bella for advice for years now. Never got accepted.

Last week, Bella finally featured my problem.

I followed her advice.”

“What was her advice?”

“The only way out of this mess is to kill my husband. Nothing else. Just kill him. If I want to live my life in peace, even inside a prison, that life would be better than what I am going through now.”

Arundhati wanted to gasp. But didn’t. She repeated slowly, “Bella… told… you… to…kill…your…..husband…..and….go….to…prison?”

“Yes.”

“And….you…..followed….that….advice…..to…..a…..T?”

“Yes.”

“Why? You could have run away. You didn’t even need to kill. You could have divorced the fellow!”

“I tried the divorce route. You, madam, know what hell it is for a woman seeking divorce from a man who refuses. You make women go from family counselling to father judges who advise us to “adjust”. You never have that advice for the husband. In the meantime, he would come and threaten me. Create public nuisance until no friend would have me in her house. No owner would rent a property to me.

You said I could have run away after killing him. But I could not have – for two reasons. One, you would have found me. Two, I don’t want to live my life like a fugitive. I want to use my own documents and be my own self.”

Arundhati had the presence of mind to show absolutely no emotional reaction whatsoever. She got a glass of water for Swapna to drink.

A police psychologist was summoned. She spoke with Swapna and certified that she was in a sound mental state and was making this confession voluntarily - no duress or influence – of people or substance.

While this was not necessary, the police officers thought it best to go through this before deploying any manpower on the case and sending police personnel to a so-called crime scene.

Once the formalities were completed, Arundhati sent a team to the crime scene.

The man was dead. No doubt about it. Other than that, the house was neat and clean. There wasn’t that much to process at the crime scene, but they still took pictures and fingerprints.

The police team, led by Arundhati, took Swapna into custody, and made the case file with the chargesheet. This was an easy case.

*********************

ACP Rajeshwari Das was rocking her flexible office chair. Her face was tilted upwards and the eyes were closed. This usually happened when she was thinking her way out of a tricky problem. The “Do Not Disturb” sign outside her door was on.

But the seats in front of her were not empty. One of them was occupied by Jatin. Jatin was Swapna’s lawyer. Swapna, who had confessed and requested a speedy trial, was not granted her wish. A government lawyer was defending her. That government lawyer was Jatin.

After meeting Swapna a few times, Jatin had been convinced that he did not want this woman to suffer for a murder that was inevitable.

The poisoning was deliberate and pre-meditated, so he could not argue insanity or self-defence.

He then used that brilliant lawyer brain and filed for “influence”. The columnist called Bella had a position of influence as far as the subject (Swapna) was concerned. She had used that influence to not just encourage, but almost coerce Swapna into committing an illegal act – by saying that this was the ONLY way for her to live peacefully. Not only that, it must also be noted that prior to Bella’s “suggestion”, the idea of murder had not occurred to the person accused. It was neither her original intent, nor her idea. She had only been used as the execution device.

Therefore, the chargesheet must be modified to include the columnist who writes as Bella as the primary accused. Swapna was under undue influence at the time of committing the murder. Further, Bella materially coerced the subject, Swapna, by indicating that this was "The only way” for her to live peacefully.

Arundhati had forwarded the submission to ACP Das. She had to admire the creativity of the young man. She wanted this case disposed of quickly. In the interest of time, she asked her assistant to reach out to Jatin and call him for a meeting.

**************

After listening to Jatin’s thoughts, ACP Das was inclined to agree with him. A battered woman anyway loses half her mind every day. Added to that was the fact that Swapna had tried every other choice – separation, divorce, police case for domestic violence, family counselling, everything. This man was making it impossible for her to lead a normal life. To such a person, the idea that murder was the only way to live peacefully, would appear to be logical and true.

And it was also true that prior to this advice, Swapna had never considered murdering Nagesh. The idea was planted in her mind for the first time by that advice columnist. The specificity of the plan had made it possible for her to execute the said plan. Both the idea and the method were given by the columnist. Swapna executed it, presumably under undue influence.

“Madam,” Jatin was still speaking, “This is the same as a woman who kills her children under the influence of some tantrik[1]. We also charge the tantrik under the same Section 302[2]. This case is identical. These advice columnists are the new tantriks.”

“Where does this advice columnist live?”

“We don’t know, ma’m. The Whois entry[3] of the website itself is also masked. We know nothing about where this Bella is from. But that alone should not stop us from making her a co-accused.”

“Jatin, the difference between the tantrik and Bella is this – we have jurisdiction over the tantrik. The main reason you are doing this is to get bail for Swapna, no? You apply for bail. I see no reason for us to contest it. She is a good person. Don’t think she will go out and do anything stupid. Let this case be. Don’t weaken it unnecessarily with this Bella nonsense. I promise to think about this some more. You have raised an important point. But in this case, I am not changing the chargesheet. Sorry.”

ACP Das closed the file with finality.

“Madam, I am not asking to add Bella only for bail. I ask because Swapna will not be the only woman going to these anonymous websites to vent. Do think about what I am saying. You are in a state that prides itself on its track record on prosecuting digital crime. Isn’t this also a form of digital crime? Inducing murder, sitting far away from the hands of the law? Do think, madam. You may not be able to prosecute this columnist. But at least you will send a message across.”

ACP Das knew, even in that moment, that the decision was not hers to make. She would have to consult her seniors.

**************

The seniors did get involved – all the way to the CM, who was the darling of the tech moguls. He dined with them at industry events, gently cajoling them to invest more money in his state – renting offices, providing jobs (creating a market for thousands of homes made and sold by his construction company).

When he heard about the case and the request of the police to file for getting the identity information of the columnist and then making her a co-accused in the trial, he kept quiet for more than a minute. Saying no would mean antagonising the police – never a good idea. Saying yes would mean raised eyebrows at the industry dinner this Wednesday.

“Let me get back to you.” He said to the state police chief.

**************

Swapna, meanwhile, had used her bail time to log back on to the site and write on the forums how she followed Bella’s column and was now accused of murder. The forums were divided – some ridiculed her for not using her judgement on something as basic as “Not killing someone”, while some others blamed Bella for suggesting something that was clearly illegal, in every part of the world. “Does Bella have a conscience at all?” Many posters asked.

The editorial team of brtdr.xyz had reached out to Bella, demanding an explanation. Now that they knew that the person had followed the advice (like everyone before her), they had to cover their backsides against any legal and user backlash. So far, they were not doing a very good job.

Bella was stunned. Her column had advised someone to kill their husband, and they had done it?

At first, she did not believe it. When the brtdr team showed her the post, she read through her own advice and was amazed at the specificity of it. The answer was worded in a very persuasive manner. Anyone reading this, even Bella, would have agreed with the answer. Now she knew why people always followed the advice given in her column.

“Is there a legal liability for me, personally?” she asked Stella, her contact from brtdr.

“My dear, the legal liability is all yours. We publish a disclaimer, see, that we are only intermediaries. The advice being sought, and given, and any liability arising out of actions therefrom, belongs completely to the two parties involved – you and the poster. Having said that, we won’t be able to avoid the backlash, I’m sure. I would suggest you get a lawyer. In the meantime, I am sending you an email, terminating our relationship with immediate effect. I’m sorry, Bella. You really were our rock star.” Stella signed off.

Bella’s head was spinning. She didn’t know what to do next.

************

Brtdr thought that once it released a statement terminating its relationship with Bella, people would move on within weeks, if not days. But the opposite happened. The user community started shrinking rapidly. Within two weeks, brtdr’s traffic was down 50%. Average active users per day dropped even more quickly – as many as 60% of the daily active users abandoned the website.

Brtdr had a business problem on its hands. The other internet forums were quick to pick it up as a ‘case study’ and discuss it threadbare.

It had no idea about the legal problem - yet.

*********************

Jatin reached out to a few media outlets on his own. Even though it was illegal to discuss sub-judice cases, he made the leak – discreetly. The journalists would report as if they had picked up the story on international forums.

The story of a columnist sitting in a first world country influencing a mentally battered person in India to commit murder, was too good to pass up. It was national news within two days. Opinion pieces flew thick and fast, and TV debates soared. Everyone was unanimous in demanding criminal action against the columnist – wherever they may be physically located.

That is how someone in Delhi read about the case. A few phone calls later, someone from the Union ministry got involved. The ambassador of the first world country was called and “encouraged” to share the details of the website owners, as a “small gesture of goodwill.”

The Ambassador made some enquiries and realised, much to his relief, that the said columnist was of Indian origin. Promptly, the personal details of the columnist were passed on to the Indian government.

************************

“So, help me understand. This is your column?”

“Yes”

“You write this advice, and you did not know that you have advised someone to commit murder, with very specific steps?”

“For the last six months, I have not been the one writing that column. I use AutoChat – an AI driven LLM[4], to write out the answers. Earlier, I would check them. But they are so good that now I don’t even read. I just copy paste the answer from the chat window to the email and send. No one else knows this.”

Jim stared. If there was a word for being speechless, he needed it now.

“Your column answers are written by AutoChat?”

“Yes. I am a computer science graduate. I learnt prompt engineering and then invested in a professional generative AI product for writing. I give the prompts for style, etc. But now the engine pretty much knows my style based on user id, and is fairly automatic.”

“So, why are you here?”

“The thing is, if I am sued, I would like to sue the makers of AutoChat for releasing a public solution before it has been tested for safety. If their product is making a kill recommendation, it has obviously not been tested for safety, right?”

“As it turns out, madam, they have gone public with the information that the solution has biases and is not tested for safety.”

“I am a professional, paying user of AutoChat. The ‘service’ being offered by AutoChat AI Solution is writing. If the writing is not reliable or even safe, what am I paying for? Why is this a commercial product?

If any liability comes to me, that is directly attributable to the work product generated by a paid product, that liability has to be passed on to the creators of the product.”

“Madam, that is how it would work in the normal course of things. But in this case, the ‘service’ you have bought is flawed, and openly so. Therefore, you assume the risk. I am sorry, there is no legal case here. I could represent you if the website sued you for a subpar or potentially damaging work product. But we could not represent you on a case against Big Tech.”

A sly smile played on Bella’s lips. “What do you think will happen if I go public with the fact that this advice column was written by AutoChat? I have lost my job already. What’s to lose? It is one thing to say a product has not been tested fully for biases. Quite another to cause a death. Do you think enterprise customers would still pay for integrating AutoChat as the base LLM in their own enterprise chatbots? After all, how much editing oversight can an enterprise provide in its own chatbot? What if the solution recommends to a user that the only way to deal with an abusive boss is to kill them?”

Jim stared, for the second time, at the woman sitting in front of him. This time, his jaw openly dropped.

“Madam, I am a corporate lawyer. I have no expertise in this kind of case. Maybe you should speak to our criminal law team. I will have a partner sent out shortly, if you would just wait here for some time.”

And Jim exited that room as quickly as he could. Any partner in the criminal team, he knew, would be more than willing to be in that room. An actual case on criminal liability of AI! And theirs would be one of the first firms to handle it!

****************

Unfortunately, that didn’t go as planned. The deliberations fell flat. Big Tech refused to assume any liability. No settlement was offered.

But someone (possibly from the law firm, but who knows?) leaked to an online AI news website the details of the case. They reached out to Bella immediately, offering her some money to share her story with them.

In tears, Bella came on video and shared how she was one of the first adopters of new tech. She had become a paying customer of AutoChat as soon as a paid version was released. She told the story as is, blaming AutoChat for the murder and for her own loss of livelihood. “My only fault is that I trusted the solution.  A solution I was told to trust and a commercial solution that I was paying for. But this is not about me. This is about what the future of humanity will be if we allow this to go unchecked. I did not provide human oversight. But this AI solution will soon form the backbone of hundreds of chatbot solutions in corporate America and global corporations. What human oversight will they provide to each answer going out of the solution? Who is responsible if this AI solution tells an employee that the only way out of their predicament is suicide? Or murder? Or mass shooting? After my experience, that doesn’t sound so impossible, does it? Their solution for an abusive husband is murder. Can you imagine the impact of that on our American society? On American families?”

The response was immediate. And widespread. Corporate America shut down every single chatbot project that included this LLM solution. Other LLM solutions came under the scanner too. “What if?” became the national question.

Like the flu, it spread – first to Canada, then Europe, then South East Asia. Pretty soon, the lawyer who had offered no settlement to Bella did not look so smart to his employers any more.

*****************

The murder case ran on in India. Jatin placed the facts of the case again in front of the judge. The sessions court decided in a record time of four months. Swapna was given life imprisonment, but parole application could be allowed in due course.

Jatin convinced her that she should be out in two, at most three years. Swapna looked at him and smiled, “And go where, Jatin ji? Who is waiting for me outside? No house, no money, no family. This is my life now. I am happy.”

But it was not the prison sentence that made headlines. It was the quote in the judgement.

“We are possibly looking at the first crime attributable to the influence of AI. This should serve as a strict warning to all of us in general, but the architects of AI in particular. AI optimises outcomes in a rational, inhuman manner. What is to stop AI from recommending mass culling of humans in the event of a drought, famine, or, say, a global pandemic? That is the mathematically optimal solution, but is it the right one?”

**********************************

The US Congress invited the CEOs of the top four AI companies for a Congressional hearing.

The CEOs admitted that the product probably should not have been offered as a commercial solution just yet, but denied any deliberate wrongdoing. They maintained that they acted ‘to the best of their ability’ but were not in a position to provide any guarantees in the foreseeable future.

All commercial AI LLM solutions that provided chat output were immediately taken off the market.

Big Tech was not happy, of course. Every day without revenue is a loss-making day. More importantly, LLMs were, so far, free of any regulation. This Congressional hearing meant that future LLM AI solutions were likely to face some sort of standards requirement.

Brtdr.xyz received a legal notice.

Instead of trying to fight the Big Tech behemoth, they decided to shut shop.

Brtdr closed the website a month after receiving the legal notice. All employees were paid the highest severance the company could manage – the CEO personally saw to that. Thousands of users who were still loyal, wrote heartfelt posts about what the website had meant to them.

In her house, Bela Ravindran sat with a friend. Two cups of coffee were on the table, and there was silence.

“Say something, Bela. Anything. What comes to your mind first when you think of everything that’s happened? Is it your job loss? How will you earn again? What will your family think of you? That poor woman? What is it? Just tell me!”

“Only this, Sharada – Who, really, is responsible for that man’s death? His wife was under undue influence. I relied on a professional writing service. And the creators of that AI? Who, really?”

 


[1][1] Tantrik – Magic healer

[2] In Indian criminal law, Section 302 applies to murder.

[3] Whois entry tells us who owns a website, their address, etc.

[4] LLM – Large Language Model. These AI programs are fed on a vast amount of conversational data and are then used to create chatbots like ChatGPT and others.