Newsletter

Obizworks AI Brief #3

By Obizworks Editorial · 2026-11-06 · 5 min read
🛈 Governed AI, human-reviewed. Drafted by Obizworks' governed AI and reviewed by a human before publication.

Welcome to the Obizworks AI Brief: one short email a week on what changed in AI, what it means for a small or mid-sized business, and a few things you can try. This issue covers 25 October to 1 November.

The big picture

Capable AI got cheaper again this week, and one of the biggest AI companies asked to go public. At the same time, the security news showed that the keys to your AI tools are now worth stealing. The lesson: enjoy the lower prices, but treat every AI key and every connected app like a bank card.

What happened

1. Anthropic files to go public. On 26 October Anthropic, the maker of Claude, announced plans to list on the Nasdaq under the ticker ANT. Its filing shows $16.3 billion of revenue in the first half of 2026. Most of its $80.7 billion net loss came from accounting changes on financing instruments, not from running the business. Why it matters: a public company has to report its numbers every quarter. For businesses that rely on Claude, that means more visibility into how stable the supplier is.

2. OpenAI halves the price of a strong model. On 26 October OpenAI launched GPT-6.1 Terra at $1 per million input tokens and $5 per million output tokens, half the price of GPT-6.1 Sol. It is built for computer use, meaning an agent that clicks through screens and forms, and OpenAI says it finishes tasks more than three times faster than its previous models. Why it matters: if you tested an AI agent and found it too slow or too expensive, test it again.

3. A strong Chinese model at a low price, and open soon. On 23 October Moonshot AI released Kimi K3.1 Pro through its API at $2 per million input tokens and $10 per million output tokens, with a very long memory of 1 million tokens. Moonshot plans to release the model weights on 4 November, so anyone could run it on their own servers. Why it matters: more choice keeps prices falling. Before you use any model with customer data, check where the data is processed and who can see it.

4. Google gives its newest model to security teams first. On 28 October Google announced Gemini 4 Carbon. It is being released in stages, starting with "trusted cyber defenders", because it is very good at finding and fixing security holes. Businesses and consumers get it later. Why it matters: the AI makers now treat security as the first job for their best models. Your IT provider should be using these tools for defence.

5. A popular AI tool shipped a version that stole keys. On 21 October one release of LiteLLM, a widely used tool that connects apps to AI models, was hijacked for a few hours and published with code that stole passwords, cloud logins and AI keys. The advice for anyone who installed it: assume the keys were taken and replace them all. Why it matters: a stolen AI key lets someone else run AI on your bill. Know which tools hold your keys.

6. Microsoft offers buyouts to up to 60,000 staff. On 26 October Microsoft offered voluntary exit packages to up to 60,000 employees in the US, UK, Canada and India. Microsoft did not link the offer to AI, but it is one more sign that large companies are reshaping their teams. Why it matters: when big firms change how they staff, small firms can hire experienced people, and should look at which roles AI can support.

From the podcasts

Nate B Jones, AI News & Strategy Daily, "Black October Isn't Just A PHP Bug. Attackers Want Your AI Keys." (24 Oct). Nate's point is simple: AI keys now behave like cash. If someone steals the key you pasted into an app, they can use AI on your account and you get the bill. He also covered a certificate company whose secret keys were so weak that one of them was the number 4. His practical advice: open the "connected apps" page of your Google and Microsoft accounts, read what each app can do, and remove the ones you no longer use.

Moonshots with Peter Diamandis, "Anthropic's $2T Cash Machine, the End of Neoclouds, and Chinese Models Catch Up" (#306, 28 Oct). The panel discussed the Anthropic listing, a sharp fall in the share prices of companies that borrowed money to buy AI chips and rent them out, and how fast Chinese models are catching up. Our takeaway: prices for strong models keep falling, so avoid long contracts that lock you into today's prices.

Moonshots, "AI's Stock Market Fallout, Google's Profit Problem, and Altman's 'Hold My Beer' Moment" (#307, 31 Oct). The panel discussed a very large crypto exchange theft that, they said, started with one engineer's computer being taken over. Their point: the weakest link is usually a person, not the technology. Train your team to spot fake messages and never share login codes.

Three things to try this week

1. Clean up your connected apps. In your Google and Microsoft accounts, open the list of apps that can access your data. Remove anything you don't use.

2. Make a list of your AI keys. Write down every AI service your business pays for, where its key is stored, and who can see it. Replace any key that has ever been pasted into a chat or an email.

3. Re-test one AI task with a cheaper model. Pick one job your team does with AI and run it on a newer, cheaper model. Compare the quality and the cost.

On the Obizworks blog

Sources