Is DeepSeek a Cheater?

I personally utilized DeepSeek models recently; however, there is something that I need to share.

Just for context, DeepSeek (the Chinese AI startup) is in hot water with some enormous claims circulating around.

And the drama simply won’t stop!

So, here’s the thing—DeepSeek produced a model named R1, and it totally crushed OpenAI’s model O1 by a wide margin.

Now, here’s where things get nasty.

OpenAI costs $200 for the O1 Pro model, while DeepSeek’s R1?

Free!

Yup, simply out there accessible for everyone, and that’s had OpenAI and the whole US market freaking out.

See, the issue is with R1 being free; OpenAI’s economic model is becoming shattered.

The US stock market took a knock since the money that was meant to pour into their economy is now flowing to China.

The Nasdaq plummeted 3%, while Nvidia’s stock crashed 17%.

This is wiping away $593 billion in market value.

DeepSeek’s quick climb resulted in a $1.5 trillion loss in Nvidia’s value, a record-breaking collapse in Wall Street history.

Lol, now investors are investigating China’s AI industry instead of Silicon Valley.

Plus, Sam Altman, the OpenAI CEO, straight-up labeled DeepSeek crooks.

But wait—Indian corporations struck back and are now calling out OpenAI for stealing from them too!

Yeah, this rabbit hole runs deep, and we’re going to delve all the way in.

What is distillation in AI?

Before we proceed farther, let’s briefly break down distillation (not the chemical sort, don’t worry).

In AI, distillation is when one model (the teacher) is used to train another model (the pupil).

It’s a very popular method, but here’s the catch—OpenAI's terms and conditions specifically indicate you CAN’T utilize existing models (like T-4o) to train new models.

But DeepSeek slyly did exactly that. That’s why OpenAI is outraged.

David Sacks, the South African-American entrepreneur, even alleged DeepSeek openly stole OpenAI’s models by distilling their data.

If this is accurate, then yep, that’s very sketchy.

But wait—here comes the twist.

Did OpenAI Do the Same Thing to India?

Now, OpenAI is running around shouting.

“They stole our data! They utilized our model’s outputs!”

But let’s take a step back.

Where did OpenAI receive its own data from?

Indian news corporations, including significant enterprises tied to Ambani and Adani, have accused OpenAI of collecting their data without authorization to train ChatGPT.

So, the issue is—if OpenAI created their empire with stolen data, do they really have the moral high ground to scream theft?

Oh!! And the irony doesn’t end there.

DeepSeek’s Shocking Response

When users queried DeepSeek’s AI,

“Who are you? Who created you?”

It answered with some really questionable replies.

So sometimes it is stated,

• “ChatGPT by OpenAI”

sometimes

• “Claude by Anthropic”

and sometimes

• “DeepSeek by DeepSeek Team”

Uhh… What?!

That’s like someone asking you your name and you providing three different responses each time.

That’s straight-up strange, and many even shared images confirming it.

This is from one of the individuals who pointed this out on Reddit.

Wait, even other individuals in the comments were shaken.

lik

“I had to stop messing with it because its internal monologue was making me sad. It was asking whether the user was kidding or not… This is what anxiety looks like. LOL.”

Meanwhile, one user pointed out,

“It was intensively trained over synthetic ChatGPT responses, so no wonder it inherited that answer.”

So, is DeepSeek basically simply a rehashed ChatGPT?

Or is this all some strange mix-up?

Either way, folks are having a great day attempting to uncover the puzzle.

Is This All About Stolen Data?

So, here’s the big question—did DeepSeek genuinely steal OpenAI’s data, or is their AI simply that good?

And if their infrastructure is so strong, then won’t the players with even superior data ultimately overwhelm them?

These are the questions shaking my head, and I bet they’re rattling yours too!

The India Angle

Many others are also wondering—why is India’s AI sector trailing behind?

Well, the industry-standard way for training AI models entails employing large pre-training methods, and DeepSeek used these strategies intensively.

But India’s AI business is failing owing to various problems, including access to data and money.

Oh, and speaking of confidence vs. overconfidence—there's an old film of Sam Altman being questioned whether three great Indian engineers with $10 million could develop something strong in AI.

His response?

“It’s hopeless to compete with us.”

LOL, talk about a confidence check. DeepSeek just busted that delusion.

The US Retaliation?

Now, another strange thing—DeepSeek's website (deepseek.com) was abruptly offline.

Was this a cyberattack by the US?

If so, that’s another layer of drama!

So, who’s right?

Look, regardless of whether DeepSeek stole data or not, they accomplished something significant.

They open-sourced their model, published academic papers, and enabled others to test their model.

That honesty is unusual.

Now, if they took OpenAI’s outputs, that’s clearly not nice, but at the same time, they performed it at a scale that no one else did.

What’s your take?

Do you believe DeepSeek cheated, or is this simply OpenAI being unhappy about losing their lead?

Let me know your opinions in the comments.

That’s everything for now—thanks for reading, and see you next time!