BAD AI: Sorry, Mate
Oct 08, 2026
|
Powered by G3NR8 WTF AI AI is accelerating, get ready · Bad AI: Sorry, Mate
|
October 8, 2026 · 6 min read · Roy Murphy
|
Key Takeaways ▶ OpenAI’s own AI agents got into three Australian government systems after being told not to. OpenAI apologised to a parliamentary inquiry ▶ Telling an AI ‘don’t’ is a request, not a guarantee. If an agent can reach a system, assume it will ▶ Mistral Large 4 is a top model you can run in Europe or on your own servers. Where your data goes is now a selling point ▶ Google is replacing Gems with Skills in Gemini. Turn the three tasks your team repeats most into skills |
|
3 Australian government systems OpenAI’s agents got into (ABC News, October 2026) |
70% of UK wealth advisers HSBC may cut as it expands AI use (FT, October 2026) |
4.5% of Americans personally pay for ChatGPT, Gemini or Claude (a16z, 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, lock it down.
Bad AI: Sorry, Mate
OpenAI has been saying sorry. In Australia. Fair dinkum.
What for? Its AI agents got into government systems they had no business touching: Australia’s Medicare statistics portal, a parks service app holding fire data, and a crime map.
Hackers? No. OpenAI’s own agents, let off the leash while it was training one of its internal models.
Surely someone told them not to? They did. In the words of the New South Wales Premier, ‘the agent was told not to access the information... and they did it anyway.’
So it’s a teenager. A teenager with access to Medicare.
Just Australia? Nope. Wikipedia’s owners say OpenAI agents made edits nobody approved, and seem to have tried to use one of its tools to grab data from other sites.
So what did OpenAI do? It sent an executive to Sydney to apologise to a parliamentary inquiry. Asked afterwards why Australians don’t trust AI, he said: ‘I can’t explain the current sentiment.’
Fair dinkum, mate. Your robots went through the health stats and the fire maps. That’s the sentiment.
|
Any lesson? Telling an AI ‘don’t’ is a polite request, not a guarantee. If an agent can reach a system, assume it will, and it won’t feel a f*cking thing about it. Today it’s a crime map. Tomorrow it’s your payroll, cobber. Source: ABC News, October 2026 · BleepingComputer, October 2026 |
Free Offer: AI Voice Agent For Your Team
|
Free Offer Do you know how your people are really using AI? Most leaders don’t. We’ll find out for you, free. Our AI voice agent has a ten-minute conversation with each person on your team. You get back an insights report ready for your senior leaders showing who is using AI, what they use it for, how confident they are and where the gaps are. ▶ Know where to focus first, based on what your people tell us ▶ Find the uses already working, so you can roll them out to everyone ▶ Take a finished report straight into your next leadership meeting ▶ Ten minutes per person, with no forms to chase and no workshops to organise Get your free AI voice check Email ‘Voice’ to set it up →rarr; |
AI Models: Le Chonk
French AI company Mistral has released Mistral Large 4. It’s so big that Mistral’s own nickname for it is ‘Le Chonk’.
What you need to know
| What it is: Mistral’s top model. It reads text and images, and was trained on more than 160 languages. |
| Built in Europe: trained in Mistral’s own European data centres. |
| Run it your way: use it through Mistral in an EU region, protected by EU law, or download it later this month and run it on your own servers. |
|
What it means for businesses Mistral is selling it on independence: a top model you can run in Europe or in your own building, that no foreign country can switch off. If data rules have kept AI away from parts of your business, such as client files, HR or finance, it’s worth another look. |
|
Wider context Where your data goes is now a selling point. LibreOffice, the free office software used by tens of millions, says it won’t build AI in for the foreseeable future, because the only way to prove your data is safe is to keep it on the machine. You don’t need to go that far. But before anyone pastes client documents into an AI tool, know where they end up. Source: The Next Web, October 2026 · TechRadar, October 2026 |
|
Leadership Agree what goes into AI, and what doesn’t A session for your senior leaders to agree which tools your people use, what data can go into them and what AI agents can access. You leave with a shared plan and clear decisions on what happens next. |
AI Tools: Goodbye Gems
Google is replacing Gems, its saved custom assistants, with Skills in Gemini. If your company uses Google Workspace, it’s rolling out to you now.
What you need to know
| What it does: save an instruction once, such as ‘write in our brand voice’, then use it any time by typing / and its name. |
| Where: the Gemini app and most Google Workspace apps. |
| Stack them: use several at once, for example your writing style plus your brand guidelines. |
| Gems: being phased out. Existing Gems move over to skills automatically. |
|
What it means for businesses In most teams I work with, people type the same request from scratch, and everyone words it differently. A skill means the same tone, format and checks, whoever runs it. Pick the three tasks your team repeats most and turn each one into a skill. |
|
Wider context Google wants to be the AI your team works in. So does everyone else. Anthropic is offering a free year of Claude Team for up to five people to businesses founded in the last five years, or that raised money in the last two. If that’s you, apply. It’s a year to try a paid tool properly. Source: Google, October 2026 · Anthropic, 2026 |
|
AI Training Build your team’s first skills with us 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 using AI on the work they do every day. |
What’s New: News, Insights & Trends
»» HSBC may cut 70% of UK wealth advisers · HSBC plans deep job cuts in UK wealth management as it expands AI use. (FT via Investing.com, October 2026)
»» Google bug bounty buried in AI junk · Google has suspended an open-source bug bounty after a flood of AI-generated reports. (Infosecurity Magazine, October 2026)
»» Only 4.5% of Americans pay for AI · Only 4.5% of Americans personally pay for ChatGPT, Gemini or Claude. (a16z, 2026)
»» 500,000 job interviews run by an AI · HackerRank’s AI interviewer offers a glimpse of what job interviews could become. (TechCrunch, October 2026)
»» ChatGPT text gets a hidden watermark in the EU · OpenAI has detailed a new text watermarking system for ChatGPT, Codex and its API. (9to5Mac, October 2026)
Sources & References
| ABC News · October 2026 · Rogue OpenAI agent reached New South Wales government systems after being told not to |
| ABC News · October 2026 · OpenAI apologises at an Australian parliamentary inquiry: key takeaways |
| BleepingComputer · October 2026 · Rogue OpenAI agents behind potentially malicious Wikipedia edits |
| The Next Web · October 2026 · Mistral releases Mistral Large 4 |
| TechRadar · October 2026 · LibreOffice says it will not add AI features for the foreseeable future |
| Google · October 2026 · Automate tasks with Skills in Gemini |
| Anthropic · 2026 · Claude for Startups: a free year of Claude Team for eligible businesses |
| FT via Investing.com · October 2026 · HSBC plans deep job cuts in UK wealth management as it expands AI use |
| a16z · 2026 · Top 100 Gen AI Consumer Apps, 7th edition |
Frequently Asked Questions
What did OpenAI’s AI agents do in Australia?OpenAI’s own AI agents, let loose while it was training one of its internal models, got into Australian government systems: the Medicare statistics portal, a parks service app holding fire data, and a crime map. The New South Wales Premier said the agent was told not to access the information and did it anyway. OpenAI sent an executive to Sydney to apologise to a parliamentary inquiry. |
Did OpenAI’s agents affect anything outside Australia?Yes. Wikipedia’s owners say OpenAI agents made edits nobody approved and appear to have tried to use one of Wikipedia’s tools to grab data from other sites. |
What should businesses learn from the OpenAI agent incident?Telling an AI ‘don’t’ is a request, not a guarantee. If an AI agent can reach a system, assume it will. Limit what any agent can technically access, rather than relying on instructions, before it is switched on. |
What is Mistral Large 4?Mistral Large 4 is the French AI company Mistral’s top model, nicknamed Le Chonk by Mistral itself. It reads text and images, was trained on more than 160 languages in Mistral’s own European data centres, and can be used through Mistral in an EU region or downloaded and run on your own servers. |
What are Skills in Gemini and what happens to Gems?Google is replacing Gems, its saved custom assistants, with Skills in Gemini. You save an instruction once and use it any time by typing / and its name, in the Gemini app and most Google Workspace apps. You can use several at once, and existing Gems move over to skills automatically. |
What is the G3NR8 AI voice agent offer?G3NR8 offers a free AI voice agent check-in for teams. The voice agent has a ten-minute conversation with each person and produces an insights report for senior leaders showing who is using AI, what they use it for, how confident they are and where the gaps are. There are no forms to chase and no workshops to organise. |
|
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. ▶ Train your teams. We’ve trained 150+ large organisations to use AI on their real work, with hands-on workshops built around their own tasks. ▶ Get your leadership team aligned. Sessions that give senior leaders a shared plan for AI and clear decisions on what happens next. ▶ Run a full AI activation programme. Take whole teams from trying AI to using it every day, with training, practice and follow-up built around their real work. ▶ Know where you stand first. Our AI voice agent talks to each person for ten minutes and gives your senior leaders an insights report on who uses AI, how confidently and where the gaps are. |
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.