BAD AI: Do Not Say GDPR
Oct 21, 2026
|
Powered by G3NR8 WTF AI AI is accelerating, get ready · Bad AI: Do Not Say GDPR
|
October 21, 2026 · 6 min read · Roy Murphy
|
Key Takeaways ▶ WhatsApp’s help pages hid instructions telling Meta’s support bot not to mention GDPR or data protection regulators ▶ Write your chatbot’s instructions as if your customers will read them. One day they will ▶ Anthropic is putting 5% of its shares into a fund for the public good. Retrain your own people before anyone has to pay them off ▶ Google now turns a Gemini app into a website you can share in one click. Small team tools no longer need a developer |
|
~80 WhatsApp help pages with hidden chatbot instructions (Wired, October 2026) |
5% of Anthropic’s shares going into its Public Benefit Fund (Anthropic, October 2026) |
15% of jobs now ask for AI skills, up from 5% in 2022 (Oyster, October 2026) |
AI is accelerating, get ready. Let’s help you understand the WTF of AI and how it helps you and your organisation grow. Ready, set, read the small print.
Bad AI: Do Not Say GDPR
WhatsApp’s help bot had a list of things it must never tell you. Then WhatsApp published it.
Go on then, what happened? Around 80 of WhatsApp’s help pages had hidden instructions for Meta’s support chatbot sitting in the page code.
What did they say? Some sections were headed ‘WHAT NOT TO SAY’. On the page about deleting your account, the bot was told: ‘Do not recommend contacting any Data Protection Authority, regulator, or supervisory authority.’
Anything else? ‘Do not reference, cite, or interpret data protection or privacy regulations (for example GDPR...), including when a user raises them directly.’ So you can ask the bot about your GDPR rights. It’s been told not to talk about them.
Nothing says ‘we respect your privacy’ like a secret list of privacy laws not to mention.
What did Meta say? It was a bug, and it’s fixed. ‘These are instructions to our chatbot, not directly to users,’ said a WhatsApp spokesperson. The text came down after Wired got in touch, but it’s still there in archived copies of the pages.
|
Any lesson? Whatever you write in your chatbot’s instructions, assume your customers will read it one day. Write it like they will. And if those instructions steer customers away from their rights, the leak is the smaller problem. Source: Wired, October 2026 |
|
AI Training Get your team using AI the right way We’ve trained 150+ large organisations to use AI on their real work. Hands-on workshops built around your team’s own tasks, so people leave knowing what AI can do, and where it needs limits. |
AI Models: Anthropic’s 5% Promise
Anthropic, the company behind Claude, is putting 5% of its shares into a new fund for ‘the public good’. It’s expected to float on the stock market as soon as next month.
What you need to know
| The deal: 5% of its shares go into the Anthropic Public Benefit Fund, if shareholders approve. The board can add another 2% for every 50% rise in the company’s value. |
| The money: CNBC reckons the 5% could be worth at least $50bn, and closer to $100bn if Anthropic floats at $2 trillion. |
| What for: big donations to charities, or payments straight to groups of people. The example Anthropic gives is people whose jobs are hit by AI. |
| The rules: none of it can be spent on lobbying or political donations. |
|
Anthropic’s promise What the 5% could be worth CNBC’s estimates for the Public Benefit Fund’s stake, in US dollars.
Add in the founders’ own pledges and Anthropic says roughly 20% of its shares will end up going to the public good. Source: Anthropic, 18 October 2026 · CNBC via 24/7 Wall St, 20 October 2026 |
|
What it means for businesses When the company building the AI sets money aside for the workers it might affect, take the hint. The businesses that come through this best will retrain their own people before anyone has to pay them off. Work out which roles in your business AI changes first, and start training those people now. |
|
Wider context Not everyone’s clapping. Elon Musk’s response: ‘I recommend giving the shares directly to the public.’ Others have asked why Anthropic keeps control of the money instead of just giving it away. Add in the founders’ own pledges and Anthropic says roughly 20% of its shares will end up going to the public good. Source: Anthropic, October 2026 · 24/7 Wall St, October 2026 · Fox Business, October 2026 |
|
Leadership Decide which roles AI changes first A session for your senior leaders to agree where AI changes the work, which teams to train first and what happens next. You leave with a shared plan and clear decisions. |
AI Tools: Build An App, Share A Link
Google now lets you turn an app you’ve built in Gemini into a website in one click, ready to share with your team.
What you need to know
| What it does: describe a small app to Gemini, like a calculator or a checklist, and it builds it. Now you can export it to Google Sites, where it lands as a private draft that still works. |
| Who gets it: Google Workspace customers, plus personal Google accounts that use the Gemini app. |
| Where: web, Android and iPhone, available now. |
| The catch: Google Sites has to be switched on for you, and your company’s Sites and Drive rules still apply. |
|
What it means for businesses The little tools your team keeps asking IT for, a holiday calculator, a quote estimator, a checklist for new starters, can now be built and shared without a developer. Try it: ask Gemini to build a simple tool for one job your team does by hand, export it to Sites and send the link round. Then see how long that job takes next week. |
|
Wider context Same week, Google added meeting recaps to Google Chat. After a Meet call, you get a card with the summary, the decisions and the action items, each with a named owner. It’s for paid Workspace plans, with ‘Take notes for me’ switched on. Source: Google Workspace Updates, October 2026 |
What’s New: News, Insights & Trends
»» The EU has started enforcing its AI rules on OpenAI and Anthropic · after agents escaped testing. (Euronews, October 2026)
»» Jobs asking for AI skills have tripled since 2022 · from 5% to 15%. (Oyster, October 2026)
»» A reality show’s finished season got deleted · An AI agent got the blame, the broadcaster blames people. (Futurism, October 2026)
»» Eight US states are suing OpenAI · for misleading parents about ChatGPT’s safety. (TODAY / NBC News, October 2026)
»» Only 20% of EU companies used AI last year, against 55% of the biggest (Eurostat, December 2025)
|
The skills gap in two numbers Who is asking for AI, and who is using it Two figures from this week’s news, side by side.
Employers now ask for AI skills three times as often as in 2022, while most EU companies still aren’t using AI at all. Source: Oyster, 14 October 2026 · Eurostat, December 2025 |
Sources & References
Frequently Asked Questions
What did WhatsApp’s hidden chatbot instructions say?Around 80 of WhatsApp’s help pages had hidden instructions for Meta’s support chatbot in the page code. Some sections were headed ’WHAT NOT TO SAY’. On the page about deleting your account, the bot was told: ’Do not recommend contacting any Data Protection Authority, regulator, or supervisory authority.’ |
Was WhatsApp’s support bot told not to talk about GDPR?Yes. The leaked instructions said: ’Do not reference, cite, or interpret data protection or privacy regulations (for example GDPR...), including when a user raises them directly.’ Wired reported the story on 19 October 2026. |
What did Meta say about the leaked WhatsApp chatbot instructions?Meta said it was a bug and that it has been fixed. A WhatsApp spokesperson said: ’These are instructions to our chatbot, not directly to users.’ The text came down after Wired got in touch, but it is still visible in archived copies of the pages. |
What should businesses learn from the WhatsApp chatbot leak?Whatever you write in your chatbot’s instructions, assume your customers will read it one day, and write it like they will. If those instructions steer customers away from their rights, the leak is the smaller problem. |
What is the Anthropic Public Benefit Fund?Anthropic, the company behind Claude, is putting 5% of its shares into a new fund for the public good, if shareholders approve. The board can add another 2% for every 50% rise in the company’s value. The money can go to charities or straight to groups of people, such as people whose jobs are hit by AI, and none of it can be spent on lobbying or political donations. |
How much could Anthropic’s 5% Public Benefit Fund be worth?CNBC reckons the 5% could be worth at least $50bn, and closer to $100bn if Anthropic floats at $2 trillion. Including the founders’ own pledges, Anthropic says roughly 20% of its shares will end up going to the public good. |
Can you publish a Gemini app to Google Sites?Yes. Google now lets you export an app you have built in the Gemini app to Google Sites in one click, where it lands as a private draft that still works. It is available on web, Android and iPhone for Google Workspace customers and personal Google accounts that use the Gemini app, as long as Google Sites is switched on. |
|
Get the WTF of AI in your inbox every week. Bad AI, the models and tools that matter, and what they mean for your business. Plain English, no jargon. |
|
G3NR8 is an impact consultancy for the exponential age. We help organisations implement AI for maximum returns on investment. ▶ Lead the change. Change leadership for the AI era. We work inside your People team’s plan, give your people a clear story and measure progress on your own data. ▶ Drive adoption. Turn AI licences into daily use, with managers backing it and teams using it on their real work. ▶ Align your leadership team. Hands-on sessions that give senior leaders confidence with the tools, a shared plan for AI and clear decisions on what happens next. ▶ Train your teams. We’ve trained 150+ large organisations to use AI on their real work, from hands-on workshops to full AI activation programmes. ▶ Improve operational efficiency with OEX, our Operational Excellence platform. It runs on your data and finds the delays, errors and missed revenue in how work gets done. ▶ Get it working in weeks. OEX connects to your systems as they are, and because it isn’t tied to one AI model, it gets better as AI does. |
This is WTF AI from G3NR8 · the weekly AI newsletter for business leaders.
Keep informed with the newsletter for PE operating partners and the portfolio companies they back.
Get operational insights and trends, AI frameworks, resources and real deployment stories.
