In the World of AI: Week of 6 October 2026, ChatGPT Moves to GPT-6

ChatGPT's main model changed for everyone, Anthropic released a much cheaper small model, and a survey of Australian finance professionals showed that AI use is wide but shallow. This is the week from 7 to 13 October 2026, with one step for each story.

1. ChatGPT rolls out GPT-6 to every plan

Android Headlines reported on 7 October that OpenAI is rolling out GPT-6 in ChatGPT worldwide. Plus, Pro, Business and Enterprise users get GPT-6 Sol in the main chat tab straight away, with Enterprise timing depending on admin settings. Free and Go users get a model called GPT-6 Luna from 8 October. GPT-6 replaces the "Latest" option in the model picker, though people can switch back to GPT-5.6 Sol. An optional "Intelligent UI" lets ChatGPT add clickable diagrams and small widgets to answers, and users can turn the visual parts off in web settings. The speed claims are OpenAI's own.

Why it matters: Staff on free accounts and staff on a paid business plan are now using different models, and answers may read differently from last week's. If anyone pastes client information into a free account, the plan they use matters.

This week: Ask your team which ChatGPT plan each person uses, and move work that involves client information onto the business plan. Our answer on whether ChatGPT is safe for client data explains the difference.

2. Anthropic releases Claude Haiku 5.5

Anthropic's launch page for Claude Haiku 5.5 is dated 7 October 2026. It calls the model its cheapest, fastest and most capable small model, aimed at high-volume work such as summaries, classification and live support. Anthropic says it costs about 75% less to run than Haiku 4.5 on average. The same announcement says Team subscribers now receive a monthly API credit that works on any model.

Why it matters: For repetitive jobs, like sorting incoming enquiries or tagging invoices, the small model is usually the right size. Lower running costs can turn a marginal automation into a worthwhile one.

This week: If you pay per use for an automation that handles high volumes, ask your builder whether the smaller tier would do the job.

3. The government signals how it may regulate frontier AI

An MLex report published on 9 October says Assistant Minister for Science and Technology Andrew Charlton spoke on Thursday 8 October. He said the government is considering a systems-based approach to frontier AI, where companies would have to set up a rigorous process to identify, test, report on and manage risks, and be accountable for whether it works. He argued voluntary codes do not counter competitive pressure and that prescriptive rules could date quickly. The report does not say whether small businesses would be covered or how the approach would be enforced.

Why it matters: The direction is toward proof that a process exists, not a list of banned uses. That pattern tends to flow down the supply chain, so suppliers may start asking customers how they manage AI risk.

This week: Name one person who owns AI decisions in your business. The AI policy answer shows what a short policy covers.

4. CPA Australia finds wide AI use and shallow adoption

The Accountant reported on 9 October on CPA Australia's 2026 Business Technology Report, which surveyed 1,679 accounting and finance professionals across Australia, Hong Kong, Malaysia, the Chinese Mainland and Singapore. It found 88% of Australian businesses use AI but only 10% have embedded it extensively, the lowest rate of the markets surveyed. Some 28% use agentic AI, meaning tools that complete multi-step tasks, and almost three in ten organisations give staff little or no guidance on using AI. Barriers named included privacy, cybersecurity and weaker human oversight.

Why it matters: The gap between 88% using AI and 10% embedding it suggests most businesses are experimenting without a settled process, and the survey's own barriers list points to privacy and oversight as the sticking points.

This week: Write one page telling staff which tools are approved, what data must never go in, and who to ask. Share it with the team.

5. Gemini 4 Argon reaches general availability

Google's post of 13 October says Gemini 4 Argon is now available globally to developers and enterprises through AI Studio and the Gemini platform. Google AI Ultra and Pro subscribers can use it in the Gemini app and in AI Mode in Search, and Ultra subscribers can use it in Google Antigravity. Introductory API pricing ends on 30 November 2026, after which the rate rises. This follows the limited release of 30 September.

Why it matters: If your business runs on Google Workspace, you will start to see Gemini features improve, and anything built on the API faces a price change at the end of November.

This week: If a developer built a tool on Gemini for you, ask whether the 30 November price change affects your bill.

What to do this week

• Find out which ChatGPT plan each staff member uses, and move client work to the business plan.
• Ask your builder whether a small model could cut the running cost of a high-volume automation.
• Name one owner for AI decisions.
• Circulate a one-page rule on approved tools and what data stays out.
• Check whether any Gemini-based tool is affected by the 30 November pricing change.

Sources

• Android Headlines, "OpenAI bringing GPT-6 AI models to ChatGPT for everyone", 7 October 2026: androidheadlines.com
• Anthropic, "Claude Haiku 5.5", 7 October 2026: anthropic.com
• MLex, "Australia looks to systems-based regulation for frontier AI standards", 9 October 2026: mlex.com
• The Accountant, "Australian businesses at risk of falling behind in AI race, CPA Australia finds", 9 October 2026: theaccountant-online.com
• Google, "Gemini 4 Argon: Built to solve your hardest problems", 13 October 2026: blog.google