BAD AI: Deloitted 2

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, agent.

This week:

😱 Bad AI: Deloitted 2

🔥 AI Offer: Marketing AI Lunch & Learn

✨ AI Models: DeepSeek V3.2

💭 AI Battles: OpenAI Code Red

💡 News, Trends & Insights

BAD AI: Deloitted 2

Deloitte has had quite the week.

Big client win? Award recognition? Not quite. AI has allegedly been making stuff up for one of its government clients.

Didn't something similar happen recently? Yea, the Aussie government AI caper. This was another alleged colossal f*ck up minor referencing error.

Jesus wept, what's happened this time? Similar accusations of adding false citations, including data from made-up academic papers and creating a tappable 3D portal to Narnia on page 12*.

What's the reaction been like? Almost 100% positive - AI Hallucination Weekly called it 'a turning point in evidence-optional consulting' with Nobells prize research winner Dr Klaus Nonsense-Metric, adding 'as someone who lives only in fictional footnotes, I applaud their efforts to upgrade my influence'.

You just made that up? How dare you. I simply leveraged cross-functional imagination assets to fill an inconvenient evidence gap.

Wonderful, any other possibly unrealistic elements worth mentioning? Probably the invoice.

AI LEARNING: AI Lunch & Learn Session

We're offering a free online AI Lunch & Learn Session for your marketing team.

In 45 minutes, we'll show you how real teams are using AI to scale content, improve workflows and get more done with less.

✅ The best use cases for AI
✅ AI workflows with examples
✅ Impact and opportunity areas
✅ Q&A to answer any questions

Just reply to this email with 'Lunch' to set up a session for your team.

AI MODELS: DeepSeek V3.2

Chinese AI lab DeepSeek has released V3.2 and V3.2-Speciale, two reasoning models matching GPT-5 and Gemini 3 Pro on key benchmarks - and they are open-source.

• Reasoning first: The models are built to improve multi-step reasoning, structured problem solving and agent-style workflows - DeepSeek is positioning them as reasoning models rather than general chatbots.
• Open weights, open questions: They ship under MIT licence on Hugging Face - but being China-based raises data handling, compliance and geopolitical considerations for enterprise use.
• Cost: DeepSeek's API pricing is stated to be lower than leading proprietary models (GPT, Gemini, Claude), making it potentially useful for experimentation.

What it means for businesses:
The performance vs cost is worth reviewing, but proceed with caution. There are risks around data residency, regulatory compliance and long-term access if US-China tensions escalate.

Wider context:
DeepSeek's R1 release earlier this year rattled markets and sparked chip export control debates. V3.2 proves it wasn't a fluke - but whether Western enterprises can practically adopt Chinese AI infrastructure remains an open question.

AI BATTLE: OpenAI Code Red

OpenAI CEO Sam Altman has declared a code red to fast-track ChatGPT improvements, while a new model codenamed 'Garlic' is being developed to counter gains from Google and Anthropic.

• Internal urgency escalation: It's a directive to push resources towards improving ChatGPT's speed, reliability and personalisation - potentially pausing work on other initiatives.
• New model in development: OpenAI is training 'Garlic' - possibly launching as GPT-5.2 or GPT-5.5 by early 2026.
• Competitive pressure: Google's Gemini has increased to around 650 million monthly users while OpenAI still has a big lead with over 800 million weekly users.
• Next week's release: Altman confirmed a new reasoning model is dropping next week that internal tests show beats Gemini 3 - a push to reclaim the capability narrative before year-end.

What it means for businesses:
This is a shift in AI model dynamics. OpenAI's scramble means they're concerned that enterprise buyers have more leverage, and the gap between leading models is narrowing. If you're locked into one provider, now's a good time to test alternatives.

Wider context:
Three years ago, Google declared its own code red when ChatGPT launched. Now the tables have turned. Google's infrastructure and relentless launching of AI experiments, apps and products are starting to bite.

Talk to us. We help organisations work with the right AI workflows and models for your highest impact use cases.

What's New - News, Insights & Trends

>> BOE weighs in on 'AI bubble'

>> China overtakes US in open models

>> Anthropic non-profit play

>> New Runway video updates

>> Track Santa with AI

FAQ: What happened with Deloitte's AI hallucinations?

Deloitte has faced allegations of AI-generated content containing false citations for government clients, including made-up academic papers and fabricated data. This follows a similar incident with the Australian government where Deloitte had to issue refunds for AI hallucination errors in official reports.

FAQ: What is DeepSeek V3.2 and why does it matter?

DeepSeek V3.2 and V3.2-Speciale are open-source reasoning models from the Chinese AI lab that match GPT-5 and Gemini 3 Pro on key benchmarks. They are released under MIT licence on Hugging Face, making them freely available but raising questions about data handling and compliance for enterprise use.

FAQ: What is OpenAI's code red and what does it mean?

OpenAI CEO Sam Altman declared a code red to fast-track ChatGPT improvements in speed, reliability and personalisation. This internal urgency escalation responds to competitive pressure from Google's Gemini and includes development of a new model codenamed 'Garlic' potentially launching as GPT-5.2 or GPT-5.5 by early 2026.

FAQ: How does DeepSeek V3.2 pricing compare to other AI models?

DeepSeek's API pricing is stated to be lower than leading proprietary models like GPT, Gemini and Claude, making it potentially useful for experimentation. However, enterprise adoption carries risks around data residency, regulatory compliance and long-term access due to US-China tensions.

FAQ: What are the risks of using Chinese AI models like DeepSeek?

While DeepSeek offers competitive performance at lower cost, enterprise users should consider risks around data residency, regulatory compliance in their jurisdiction, geopolitical considerations, and potential access disruptions if US-China tensions escalate. The models are open-source but being China-based raises data handling questions.

FAQ: How many users do ChatGPT and Google Gemini have?

OpenAI's ChatGPT has over 800 million weekly users, while Google's Gemini has grown to around 650 million monthly users. The competitive gap is narrowing, which has prompted OpenAI's code red response to improve their platform.

FAQ: What new AI model is OpenAI releasing?

OpenAI confirmed a new reasoning model dropping soon that internal tests show beats Gemini 3, designed to reclaim the capability narrative. Additionally, OpenAI is training a model codenamed 'Garlic' that may launch as GPT-5.2 or GPT-5.5 by early 2026.

FAQ: What does the AI competition mean for enterprise buyers?

OpenAI's scramble indicates enterprise buyers have more leverage as the gap between leading models narrows. This is a good time for organisations to test alternatives and avoid being locked into a single provider, as competitive pressure may lead to better pricing and features across all platforms.

G3NR8 is an impact consultancy for the exponential age. We help organisations implement AI for maximum returns on investment.

Work with us to:

• Rapidly improve your sales and marketing function with an AI ecosystem - from horizon scanning to content repurposing and active campaign planning.
• Help each employee find information 60% faster with advanced AI knowledge bases - secure, and based on your business logic and data.
• Create valuable MVPs that don't sit in silos and work immediately to generate productivity gains and ROI in weeks, not months.
• Train your teams - we've trained 150+ large organisations in advanced generative principles and implementation.

Drop us a line for a chat about productivity gains, cost savings, growth and how AI can scale your business.

*Jokes, not facts.