Most artificial intelligence news is written for developers and investors. This weekly summary is for owners and managers who just want to know: what happened, and does it change anything for my business? Here is the AI news from the end of September and the start of October 2026, with links to the original sources.
Last updated: 3 October 2026. AI news moves fast, so check the linked sources for the latest details.
1. OpenAI launches GPT-6.1 Sol at a much lower price
OpenAI released GPT-6.1 Sol on 29 September at its DevDay event. According to The Next Web, it aims for results close to the company's larger Astra model but at about one fifth of the price per token. Standard API pricing is reported at $2 per million input tokens and $10 per million output tokens.
What it means for you: the cost of running AI inside a product keeps falling. Features that were too expensive a year ago, such as a chatbot that reads your documents or a tool that drafts replies to every enquiry, are becoming affordable for small and mid-sized companies. If a vendor quoted you a high running cost for an AI feature, it is worth asking again.
2. Google's Gemini 4 Argon targets software security
Google released Gemini 4 Argon first to security teams in its Fairwind programme. As summarised by AI Weekly, the model scored 68% on the CWE-bench v1 test and can find and fix serious security weaknesses in code on its own.
What it means for you: AI is now helping defenders find vulnerabilities, but the same kind of technology can help attackers find them too. The practical lesson for any business with software is to keep it updated, limit who has access, and have it reviewed regularly. Old, unpatched systems are the easiest target.
3. California bans AI-only firing decisions
On 30 September, California's governor signed SB 947, known as the "No Robo Bosses Act". Law firm summaries such as this one on the National Law Review explain that employers cannot rely only on automated systems for termination or discipline. A person must review the decision, and affected employees must be told that AI was involved. The rules are reported to take effect on 1 July 2027.
What it means for you: this law applies in California, not in India. But it shows where regulation is heading: people want a human accountable for important decisions. If you build or buy AI that affects hiring, credit, pricing or discipline, keep a person in the loop and keep a record of how decisions were made.
4. Smaller stories worth knowing
- AI-made research is under scrutiny. AI Weekly's roundup notes a Pew Research study that found answers generated by AI instead of real people differed from human survey responses by an average of 12 percentage points. Be careful about using AI-generated "customers" in market research.
- Watermarking AI-designed proteins. Google DeepMind published methods for marking AI-designed protein sequences, part of a wider push to label AI-generated output.
- AI in government. Anthropic made its Claude for Government product generally available to US federal and state agencies, a sign that large organisations are moving AI from pilots into daily work.
The bigger picture for small businesses
Across these stories, three themes stand out:
- AI is getting cheaper, so more everyday business uses make financial sense.
- Security matters more, because AI speeds up both defence and attack.
- Human accountability is becoming a rule, not just good practice.
You do not need to follow every headline. You do need a plan for where AI can save your team time, and a safe way to introduce it.
How SimpleApp can help
We build AI-assisted software for businesses: AI-powered applications, AI chatbots and automation of repetitive work. If you are wondering which of today's AI news actually matters for your company, tell us about your process and we will give you an honest view, including when AI is not the right answer.
You can also read how AI helps us build software faster without cutting corners.