₿ BTC — —
Ξ ETH — —
Independent Analysis · Dubai

JA's Lookout

What ChatGPT and Big Tech Do With Your Data: Privacy, Training, and the Real Risk

You can love ChatGPT and still be smart enough to treat it like a glass-walled conference room.

Because that’s basically what it is.

John Ferrell (a Silicon Valley IP attorney) put it bluntly: lots of privacy-sensitive industries — law firms, banks, tech companies — don’t trust ChatGPT with secrets, and many outright ban employees from using it for work.

Not because ChatGPT is “evil.”
Because the incentives behind data collection are bigger than your comfort.

And the second transcript — from a Business Insider segment featuring a professor of internet governance at Oxford — goes even further: the real danger isn’t just privacy loss.

It’s mass influence through prediction, and the rise of single-point-of-failure recommendation engines that can cause whole populations to make the same wrong decisions at the same time.

So let’s break down what’s happening.

ChatGPT Isn’t a Vault. It’s a Service That Records Inputs.

Ferrell explains the basics: ChatGPT is a machine learning system trained on massive amounts of text. It generates responses by predicting likely word patterns — not by “thinking” like a human.

That matters because people treat it like a private assistant.

But the reality is: when you type into ChatGPT, your prompts and the model’s responses can be stored.

Ferrell highlights that OpenAI collects data from three main buckets:

  • Account data (especially for paid users): name, business details, billing information
  • Device/connection data: device type, browser, IP address
  • Your conversations: prompts, responses, interaction logs

That third one is where people get reckless.

Because your prompt is often the most honest thing you’ve typed all day.

And if you paste sensitive data into it — client documents, internal strategies, a draft SEC filing, proprietary code — you’re not just “using AI.”

You’re exporting sensitive material into an external system.

Who Can Access It? The Policy Language Is… Conveniently Foggy.

Ferrell’s core complaint is simple: privacy policies often say data may be shared with:

  • Vendors and service providers
  • Affiliates (a vague term that can include major partners)
  • Legal entities (when required)
  • Human reviewers / trainers

Even if the intent is quality control, “human review” is the line that should snap you into attention.

Because it means: your conversation can be seen by people.

And if you’re using third-party apps “powered by AI” that connect through APIs, you’ve added another layer of risk: now you’re not just trusting one company.

You’re trusting a chain.

That’s how breaches happen. That’s how leaks happen. That’s how “internal only” becomes “public.”

The Bigger Issue Isn’t ChatGPT. It’s the Data Economy.

Now zoom out.

The Oxford professor’s point is darker — and honestly more important long-term:

Big platforms (think Netflix, Amazon, social platforms) know a lot about us. We like to believe we’re unpredictable, creative, irrational.

We’re not as special as we think.

At scale, human behavior becomes pattern-heavy. And when platforms have enough data, they don’t just predict what you’ll buy.

They can influence:

  • what you watch
  • what you believe
  • what you fear
  • how you vote
  • what you think “everyone else” thinks

We’ve already seen early versions of this with targeted misinformation and algorithmic amplification. The professor frames it as a structural issue: when decision-making gets routed through a few recommendation engines, society gains a new vulnerability.

Not a hacker vulnerability.

A system vulnerability.

The “Single Point of Failure” Problem: When Everyone Uses the Same Brain

Here’s the scariest idea in the entire transcript:

Markets stay resilient because people make different choices.
If one person makes a dumb decision, the whole system doesn’t collapse.

But if everyone listens to the same “smart assistant” and it’s flawed — biased, miscalibrated, manipulated — then everyone can make the same mistake at once.

The professor compares it to discovering a brake failure… and realizing every car has the same brake.

That’s what monopolistic recommendation engines create:

  • shared assumptions
  • shared nudges
  • shared errors
  • shared outcomes

And that becomes political power and market power merged.

This is why data concentration is not just a “privacy issue.”

It’s a governance issue.

So How Do You Use ChatGPT Without Being Naive?

Ferrell offers practical rules — and these are actually solid:

Turn off chat history / training (Data Controls)

If the platform allows disabling chat history and training, use it. Ferrell suggests this can reduce the chance of your chats being used for training, and may limit retention.

Don’t put anything in AI you wouldn’t want read publicly

His old civics-teacher rule is brutal but correct:
never write what you’d be embarrassed to see read in open court.

Apply that double to AI tools.

Use a throwaway account for sensitive questions

If you must ask something delicate, reduce the identity link. Separate email, private browsing, minimal identifying details.

Be a “listener,” not a “talker”

This is the big one: keep prompts general. Don’t paste raw documents. Don’t upload proprietary text. Don’t feed it internal memos and expect magic with zero risk.

If you need feedback on something sensitive, sanitize it:

  • remove names
  • remove dates
  • remove company identifiers
  • summarize instead of pasting

Because the easiest privacy win is: stop oversharing.

The Bottom Line

If you treat ChatGPT like a private diary, you’re gambling.

If you treat it like a powerful tool that lives in someone else’s house, you’ll use it smarter.

And if you zoom out beyond ChatGPT, the real issue isn’t one chatbot.

It’s the data-driven world we’re building where:

  • prediction becomes influence
  • influence becomes power
  • and power concentrates into fewer systems

Use AI. Absolutely.

Just don’t be sloppy.

Because the people who “don’t care about privacy” are usually the ones who lose the most when it finally matters.

More Blogs:

Share this:

Like this:

Like Loading...

Discover more from J.A Lookout

Subscribe now to keep reading and get access to the full archive.

Continue reading