100,000 ChatGPT passwords leaked online

The account details were found on the dark web with other sensitive data stolen by PC malware

Hello, and welcome to Issue 5 of The AI Writer, your weekly update on workplace AI.

This week, we delve into news that hackers have been discovered quietly stealing and selling ChatGPT logins online. More than 100,000 accounts are thought to have been compromised already.

We also look at how UK hospitals are going all-in on AI and show you how ChatGPT can get you out of an email fix. And we focus on an app that uses AI to help users work across large organisations.

This issue will take you 8 minutes 51 seconds to read in full. Too long? Read the summaries in italics in just 25 seconds instead.

Let’s dive in.

– Rob Ashton

In this issue

  • Hackers leak more than 100,000 ChatGPT passwords

  • UK’s NHS spends millions on AI healthcare

  • How to use ChatGPT to write difficult emails

  • Top productivity app adds over 100 AI functions for corporate users

  • Why we’ve created The AI Writer


Hackers dump 100,000 ChatGPT passwords for sale on dark web

Young male in hoodie and Guy Fawkes mask looking at phone

Image: Ilya Yarmouth/Unsplash

An astonishing 101,134 ChatGPT passwords were among sensitive data harvested by malware from PCs between June 2022 and May 2023. Hackers have been selling the sensitive data online.

Security experts have warned organisations to tighten up how staff use AI after discovering more than 100,000 ChatGPT passwords for sale on the dark web.

According to cybersecurity firm Group-IB, some 101,134 ChatGPT accounts were compromised between June 2022 and May 2023. The research shows an increasing number of accounts being affected, with more than 26,000 login details on sale last month.

The source appears to be malware that infects hard drives so it can steal sensitive data and feed it back to hackers. The researchers say most of the data was harvested by three well-known pieces of malicious software: Raccoon, Vidar and Redline.

Graph showing increase in number of stealer logs containing ChatGPT account details, from 74 in June 2022 to 26,802 in May 2023.

The number of logins for sale matches the rise in popularity of ChatGPT

The rise in the number of passwords for sale matches that of the popularity of ChatGPT itself. This suggests an ongoing problem, with users unaware that their details are being stolen.

OpenAI – which developed ChatGPT – told The AI Writer that it’s investigating the accounts that have been exposed. But it was quick to point out that the AI chatbot itself was not hacked, as the information had been stolen from users’ devices, not from ChatGPT servers.

‘The findings from Group-IB’s Threat Intelligence report are the result of commodity malware on people’s devices and not an OpenAI breach,’ a spokesperson said.

Indeed, the malware is probably non-selective, stealing passwords and logins for other apps at the same time, as well as credit card and bank details.

Companies at risk

But the real threat to organisations is that logins give access to chat histories if users haven’t instructed ChatGPT not to save them.

Dmitry Shestakov, head of threat intelligence at Group-IB, warned that this is a big issue if staff are feeding ChatGPT sensitive or confidential information. ‘Given that ChatGPT’s standard configuration retains all conversations, this could inadvertently offer a trove of sensitive intelligence to threat actors if they obtain account credentials,’ he said.

The report advises protecting account logins using two-factor authentication. This would reduce the threat from hackers by requiring users to confirm their login by text or with an authenticator app as well as with their password.

Yet this won’t be possible for many people. OpenAI added two-factor authentication to ChatGPT in May, but it closed enrolments less than a month later for users who had not already switched it on. The company did not say why when we asked them, although it’s thought to be down to technical difficulties with the system.

Many of the users whose accounts were affected by the leak were developers who helped test the AI chatbot before its launch in November 2022.


UK National Health Service splashes £144m ($183m) on AI

Female doctor examining scan images on computer screen.

The NHS – one of the world’s largest organisations – is spending millions on AI-assisted healthcare. Technology that halves the time to diagnose strokes is one of the early successes.

AI apps aimed at consumers have received so much coverage recently that it's been easy to miss the fact that large organisations are already investing heavily in the new technology.

Now it's been announced that UK's National Health Service, one of the biggest of all (it employs 1.5 million people), is to get a cash injection of £21 million ($27 million) to spend on AI-assisted healthcare projects.

This comes on top of the £123 million ($156 million) that the NHS has already used to fund 86 AI technologies.

The new round of funding is set aside for tools that support doctors in diagnosing and making treatment decisions for cancers, strokes and heart conditions, among others.

More than four in five (86%) stroke units already use AI to help diagnose the brain condition. The outcome for patients is strongly linked to how quickly they receive treatment, and pilot projects have shown that AI can halve diagnosis time.

The NHS is one of the world’s biggest organisations

Such successes triple the chance of patients being able to live independently following a stroke.

'Rapid assessment and treatment are of the essence,' says Dr Deb Lowe, NHS England's national clinical director for stroke medicine. 'We now have real world evidence of the benefit [of AI] for NHS patients.'

The NHS, like many state-funded bodies, runs on an often highly bureaucratic system that can seem mind boggling to outsiders (and even to many people who work in it).

Early adopter

So the news that it is riding the AI wave may come as a surprise. In fact, it has been at the forefront of using new technology in medicine for years and was an early adopter of artificial intelligence.

Some might say too early, as it's not always had a smooth ride. In 2017, it fell foul of the UK's data watchdog, which ruled that a collaboration between an NHS hospital and Google's DeepMind AI unit broke privacy laws. 

Undeterred, the NHS has since set up a whole raft of AI initiatives. Partly, this was triggered by the Covid-19 pandemic. But the free health service is also under huge pressure from spiralling healthcare costs, recruitment problems and an increase in life expectancy that's created a demographic time bomb of expensive-to-treat elderly patients.

Rapid innovation

Such pressures are forcing it to look for new ways to become more efficient. It has even set up its own 'skunk works' AI lab, modelled on Lockheed Martin's rapid innovation unit that the aviation company used to slash development times for new aircraft towards the end of the Second World War.

Steve Barclay, the UK's health minister, has committed to rolling out AI-diagnosis technology to all the country's stroke units by the end of the year.

The three biggest employers in the world are all military and so almost certain to all be using AI already.


Structure a difficult email using ChatGPT

Use this prompt to unscramble the thoughts in your head and increase the chance of a positive reply.

We've all been there. A project has stumbled and you need to send an email to fix things quickly. The trouble is, you've got such a jumble of thoughts flying around your brain that you don't know where to start. Worse, there's a lot riding on your reply, so you really need to get it right.

In these situations, you can't beat the SCRAP formula. It works perfectly with human psychology to overcome the recipient's resistance, get them nodding in agreement and maximise the chance of a positive outcome.

What to do

  1. Copy and paste this prompt into ChatGPT.

  2. Replace the points to cover with your own bullet list*.

  3. Press Enter

  4. Personalise, edit and adapt the response. 

*The list can include background, what's gone wrong, what would make it right and how you feel about the situation.

The order doesn’t matter. If you just do a brain dump of all the things whizzing around your stressed-out brain, the prompt will program ChatGPT to structure them for you.

Behave as an executive assistant skilled in writing sensitive emails. You are polite, courteous yet assertive at all times. 

Your task is to write an email for me. To protect confidentiality, use 'RECIPIENT', 'SENDER', 'ORGANISATION A', 'ORGANISATION B' etc. as placeholders where appropriate. 

I am going to give you a list of points to cover. Use the SCRAP structure to write an email that addresses the issue and specifies a way forward. SCRAP stands for Situation (what happened initially, or brief background), Complication (what went wrong/unexpected problem), Resolution (what would make it right), Action (what needs to happen next), Politeness (a polite sign-off).

Avoid passive-aggressive statements. Use the active voice by default. But use the passive voice instead of blaming the recipient for the 'complication'. Start with a positive statement.

Do not use section headings for each of the SCRAP components. For example, do NOT write 'Situation:xxx'.

Here is the list of points:

[- Deadline for completion has been brought forward]

[- Now only have three weeks to get everything ready]

[- I'm going to draw up an action list and timetable]

[- XXX will lend us two of her staff (YYY and ZZZ)]

[- We should be ok as long as we don't delay]

[- We met recently to discuss the office move.]


  • This works with ChatGPT, but GPT-4 generally gives a better response.

  • Regenerate the response if the bot doesn’t follow your instructions exactly. For instance, it may sometimes label sections as ‘Situation’, ‘Complication’ etc., despite the prompt telling it not to.

  • Edit the response carefully (as with all AI-generated writing).

  • Make sure the first line of the email contains no negative information (eg the complication or problem), as that could trigger a negative reply. Save complications for the second section of the email.


Productivity app Click-Up goes all-in on AI

Animated GIF of screen recording, showing Click-Up summarising meeting notes

The makers of the productivity app have introduced more than 100 AI functions, to help corporate users in specific roles collaborate across teams.

These days, it seems like an app isn't an app unless it's an AI app. Scratch beneath the surface, though, and you often find that the AI functionality is only skin-deep.

Too many AI add-ons also attempt to do something that humans can do quicker, better or both. Sometimes, it's hard to see what the AI is adding at all or to tell the difference between genuine AI and mere marketing.

No-one could accuse Click-Up of such tactics. The productivity app, which has almost two million users worldwide, has raised its game to a whole new level by adding not one but more than a hundred AI functions.

AI meeting notes, summaries, brainstorming, copy editing, email – you name it, and it seems to be there.

Perhaps conscious of the potential to overwhelm users, Click-Up has taken a role-based approach. This also shows it’s aiming the AI app at corporate users.

The San Diego-based tech firm says it's designed the AI-enhanced version specifically to help teams collaborate across an organisation.

The app includes dozens of prompts aimed at specific corporate roles. They include AI templates for creating a territory plan (sales managers), devising a user-testing study (product designers) or creating a statement of work (project managers).

The company also recently acquired the corporate knowledge-management app Slapdash. So a future release that automatically works with company-wide information looks likely.

The AI-enhanced version of the app costs $5 per user, per month.

The AI Writer receives no compensation from the providers of the app featured above.

Why The AI Writer?

When I founded Emphasis, 25 years ago, it was just me, a cat and a kettle. It’s since grown to become the most trusted provider of business writing training in the world and helped more than 80,000 people, from 32 countries. We’ve worked with tech giants, top 10 law firms and major financial institutions, as well as at the highest levels of government. (We’ve even sent trainers to work with clients in the Himalayas.)

I would not claim that we’re experts in everything, as we’re certainly not. But, after working with the authors of around 100,000 documents, it’s safe to say that business writing is something we do know a thing or two about. And that includes witnessing all the various ways in which organisations get it wrong.

Now, we’re at the forefront of using AI to help them get it right.

Of course, I’ve seen nothing like the AI revolution that’s hitting us. Nobody has. But I do feel a responsibility to use our experience to help people navigate this brave new world of communication bots.

So here’s the deal. You get on with your job, while we immerse ourselves in the latest developments in workplace AI. We’ll worry about keeping up so you don’t have to. Then we’ll tell you exactly what you need to know to stay ahead of the curve.

Quick poll: Which ONE topic do you most want us to cover?

Login or Subscribe to participate in polls.

Please forward this email

Feel free to forward this email to anyone who might find it useful. Each issue of The AI Writer takes days to research and write, so the more people it helps, the better.

And if someone forwarded this to you, you can grab your own, free subscription here.

You can also catch up on all our previous issues.

Until next time, have a great week.

– Rob

PS. I’m pretty active on LinkedIn if you want to connect or follow me.