AI news November 2026: what it means for your business

In short

• AI news from late October and early November 2026 is mostly about limits: caps on free Copilot Chat, budgets in Claude, ChatGPT agents that switch on by default and models being retired.
• Thirteen developments matter for businesses, from new limits in Microsoft 365 and dots in ChatGPT Enterprise to a watermark on Claude text, the end of Gemini 2.5 in the API and a first European legal opinion on chatbots and news articles.
• Three dates call for action now. If you use ChatGPT Enterprise, decide before 18 November whether dots may switch on by default. If anything runs on Gemini 2.5 through the Gemini API, move before 1 December. If you run Microsoft 365, tell your help desk that staff without a Copilot licence may hit limits.

There is AI news every week, but little of it changes anything for an organisation. From the last week of October and the first days of November, thirteen developments stand out because they affect what you pay, what is switched on by default and which integrations keep working.

The common thread: providers are drawing lines. Free use gets limits, paid use gets budgets and alerts, and agents switch on by default unless someone changes the setting. If you do not choose, you get the vendor's default. Below, for each development: what happened, what it means and where to read more. This is the first edition, published on 5 November, covering news from 23 October to 5 November. We update this overview every week. Item 7 corrects what we wrote in October about usage billing for Copilot in the Netherlands. October's news is in AI news October 2026, and all of this month's Google news is in Google AI news November 2026.

1. Claude Enterprise: chat and Cowork combined, with budgets per team

On 23 October Anthropic merged chat and Cowork into a single Claude for Enterprise customers, on web, desktop and mobile. Tasks involving code or files keep running in an isolated cloud environment, even with your laptop closed, and outbound network traffic is off by default there. Admins decide per role which models staff may use and set spending limits for the organisation, groups and individual users (Anthropic). On 28 October Anthropic added shared group budgets with alerts at 50, 75, 95 and 100 per cent, plus reports showing what each team uses Claude for and what it costs (Anthropic).

That finally lets you cap AI spend per department instead of discovering it afterwards. Anthropic itself says the reports are not meant for judging individual performance or making employment decisions. Even so, if you have a works council or employee representatives, talk to them before you use reports like these: team-level data still shows how people work. For an overview of Claude and its plans, see What is Claude?

2. GitHub Enterprise: Copilot included, usage billed separately

At GitHub Universe on 28 October, GitHub announced that GitHub Enterprise now includes access to Copilot for every licensed user, at no extra cost for the seat. No AI allowance comes with it: AI usage is billed in AI Credits from the first request. Code completions while typing keep working without a budget (GitHub). Copilot Fusion, which spreads a task across several AI models, also became generally available (GitHub).

The pattern matches Microsoft 365 Copilot: access gets cheaper, usage becomes a variable cost. Set a budget before developers start running agents, and check actual consumption after a month.

3. Subscriptions now include API credit, and Claude Max 20x is now Ultra

Since 28 October, OpenAI has included API credit with paid ChatGPT plans. Pro gets 100, 200 or 500 dollars a month depending on the tier. Business gets 20 dollars per Standard seat and 100 dollars per Premium seat, up to 500 dollars per workspace. You claim it in the billing settings and need an API organisation of your own. Unused credit expires at the end of the billing period (OpenAI). At Anthropic, a Team plan includes up to 500 dollars a month in pooled API credit. Max (100 dollars) comes with 100 dollars and the new Ultra (200 dollars, formerly Max 20x) with 200 dollars a month (Anthropic). Ultra no longer has a five-hour limit for verified accounts, but it does have a weekly limit of ten times Pro.

For a small business on ChatGPT Business or Claude Team, that is free budget to test a first integration, such as an automation that sorts incoming email. Do agree who manages the API keys. A key that sits with one employee becomes a risk the day that person leaves.

4. Google: Gemini 4 Carbon, not yet for businesses

On 28 October Google announced Gemini 4 Carbon, which it calls "our strongest model", with a focus on cybersecurity and long, multi-step tasks. It goes first to a small group of security specialists. Developers, businesses and consumers will follow "as soon as possible", with no date. In the API it costs 2 dollars per million input tokens and 10 dollars per million output tokens during the introductory period, then 4 and 20 dollars (Google).

For most organisations nothing changes yet: you cannot use it, and the comparisons come from Google itself. More on Carbon and the rest of Google's news is in Google AI news November 2026.

5. Dutch data protection authority: AI risks at debt collectors

On 29 October the Dutch Data Protection Authority (Autoriteit Persoonsgegevens, AP) published a survey of 29 debt collection agencies. Many use algorithms and AI not only for chatbots and call notes, but also to estimate how likely a debt is to be recovered, based on factors such as credit score, location and payment history. The AP sees a risk of wrong decisions, discrimination and unequal treatment, and considers it realistic that prohibited fully automated decision-making happens without agencies realising it (AP; Accountant.nl).

Even if you are not in debt collection, it is worth reading: the AP's point applies to any organisation that uses AI to make decisions about people, such as credit, recruitment or customer onboarding. For important decisions, make sure an employee is involved who can explain the outcome, and write down where in the process that happens.

6. Microsoft: limits for Copilot Chat without a licence

Microsoft is introducing limits on more Copilot Chat features for users without a Microsoft 365 Copilot licence. For organisations with 2,000 or fewer Microsoft 365 seats this started on 2 November; for larger organisations it starts on 9 November. Copilot Chat stays included with Microsoft 365. Microsoft does not publish the limits, they can vary during the day, and admins cannot view or change them. Users who reach a limit usually see an option to request or buy a licence (Microsoft).

Nothing has to change technically, but your help desk will get the questions. And if staff are allowed to buy licences themselves, a new cost can appear without anyone noticing. Which settings to check now is covered in Copilot Chat usage limits: what should you set up?

7. Microsoft: usage billing for Copilot Business later, now including the Netherlands

On 3 November Microsoft moved the start of default usage billing for new Microsoft 365 Copilot Business licences bought through a partner (CSP) to January to March 2027, instead of 1 December 2026. The default spending limit remains 4,000 Copilot Credits per user per month, and you can change it. What is new is that the revised timeline applies in all markets, including the Netherlands, Belgium, France and Germany, where Microsoft had not initially expected this to be available (Microsoft).

This corrects what we wrote in October: usage billing will also be on by default for new Business licences in the Netherlands, just later. That gives you until the first quarter of 2027 to set limits and pick a pilot group. How it works is explained in What will Microsoft Copilot cost now that agents are billed per use?

8. Google: Gemini 2.5 retires in the Gemini API on 1 December

On 2 November Google announced that from 1 December, requests to gemini-2.5-pro and gemini-2.5-flash in the Gemini API will automatically go to newer models: 2.5 Pro to Gemini 3.8 Flash and 2.5 Flash to Gemini 3.5 Flash-Lite, at those models' prices. The switch comes with breaking changes, including in function calling (Google AI Studio). Since 26 October, new paid projects also store requests by default.

This is Google's counterpart to OpenAI retiring GPT-4o. If a chatbot or automation runs on Gemini 2.5, it will run on a different model from 1 December, and possibly not without errors. What to test before then is in Gemini 2.5 retires on 1 December: what should you do?

9. Anthropic: an invisible watermark on Claude text

Anthropic now marks Claude's output in a machine-readable way, to meet Article 50 of the AI Act. Text gets an invisible watermark in its word choice, and files such as PNG and JPEG get signed provenance data based on the C2PA standard. It covers the API, the Claude apps, Claude Code, Cowork and Claude Tag. Models launched in the EU on or after 2 August 2026 do this from launch; older models are being added, and the rollout on Amazon Bedrock completes by 10 November. You can check files with the free Claude Content Checker. On some older models the watermark may be missing if a developer changes the randomness settings in the API, such as temperature (Anthropic).

After OpenAI in October, Anthropic is the second major provider to do this. For your team, little changes. Anthropic calls a detected watermark "not fully conclusive", and a missing watermark does not prove a person wrote the text. If you build Claude into your own product, Anthropic says you need to assess for yourself what Article 50 requires. More in Will AI text carry a watermark?

10. EU Court of Justice: chatbots and news articles

On 3 November Advocate General Szpunar of the Court of Justice of the EU delivered his opinion in case C-250/25, Like Company v Google. It is the Court's first case on generative AI and copyright. According to an analysis by the law firm GSK, the press publishers' right protects the way an article is written, not the news or the facts. Risk arises when a chatbot gets round a paywall, when an answer effectively replaces the article, or when the original wording is recognisable in the output (Court of Justice; GSK).

An opinion is not a judgment, and the Court does not have to follow it. If you build an AI tool that summarises other people's news or articles, such as a daily market briefing, show the source, link to the original and do not reproduce long passages. For publishers and brands it signals that AI answers pointing to their sources are becoming the legally safer route. How to get your own content into those answers is covered in What is GEO?

11. OpenAI: dots switch on by default in ChatGPT Enterprise on 18 November

On 4 November OpenAI made dots generally available for ChatGPT Enterprise, Edu and Healthcare. Dots are personal agents that keep working on a task between conversations. An admin preview runs until 18 November: admins can switch dots on now or opt out, and after that date they become available by default. Admins control which members can use dots, can pause them and can review usage. New dots support data residency, with processing in the EU or the US (OpenAI). For each workspace you set separately whether a dot may use a browser, the network, the computer or the password manager (OpenAI).

This is the only item in this overview that switches on by itself if you do nothing. Before 18 November, decide whether you want dots for everyone, for a pilot group or not yet, and what permissions a dot gets. The five questions to ask first are in Using AI agents safely. What dots are is explained in OpenAI dots: what can you do with them?

12. OpenAI: Images 2.6 adds GIFs and faster images

On 4 November OpenAI released Images 2.6 in ChatGPT, Codex and the API. According to OpenAI, images are generated 50 per cent faster, characters, products and brand details stay more consistent across edits, and you can now create animated GIFs, for example of a mascot or a product shot. It is "available globally today" on all plans. Developers choose between GPT-Image-2.6 Flare, which is faster, and Sunburst, which is more precise (OpenAI).

Marketing teams can use this straight away for social posts and concepts. Agree which images you label as AI-generated and who signs off an image before it goes out. How to set those rules for your team is covered in Using AI safely: how to write an AI policy for your team.

13. Mistral Studio: European agents on your own servers

In early November Mistral, the French AI provider, launched Studio, a platform for building agents, workflows and internal apps on Mistral Large 4. Workflows combine fixed rules, an agent's reasoning and human approval, and every step can be traced in a log. You can run both the models and the platform yourself: in your own cloud, your own private network or on your own servers. Mistral names BNP Paribas, Capgemini and Accenture among its users. No price is given (Mistral).

For organisations that do not want data processed outside their own environment, this is Mistral's second concrete step in two months, after Large 4. When a European model is worth it is covered in Mistral AI: when should you choose a European AI model?

Which dates should go in your diary this month?

• 9 November: Copilot Chat limits without a licence also apply to organisations with more than 2,000 seats (item 6).
• 10 November: Claude watermark fully rolled out on Amazon Bedrock (item 9).
• 17 November: Gems in the Gemini app are moving; for Workspace Business and Enterprise they will not disappear before 1 March 2027 (Google AI news November 2026).
• 18 November: dots on by default in ChatGPT Enterprise, Edu and Healthcare (item 11).
• 1 December: Gemini 2.5 Pro and 2.5 Flash move to newer models in the Gemini API (item 8).
• January to March 2027: usage billing on by default for new Copilot Business licences through a partner (item 7).
• 14 January 2027: GPT-4o and GPT-4.1 leave OpenAI's API (Will your AI integrations stop on 14 January?).

What does this AI news mean for your business?

When can you ignore this?

Do you mainly use AI personally, for writing and analysis, with no agents, integrations of your own or an Enterprise plan? Then little changes for you right now. At most you will notice a limit in Copilot Chat or faster images in ChatGPT. If someone builds software or automations for you, ask whether they run on Gemini 2.5, GPT-4o or GPT-4.1. If you manage a business subscription, the default settings matter most this month: whatever you do not set, the vendor sets for you.

Sources

• Anthropic, Claude Cowork and chat come together for Enterprise, 23 October 2026
• Anthropic, Smart reports, surveys and pooled budgets, 28 October 2026
• GitHub, GitHub Enterprise now includes access to GitHub Copilot, 28 October 2026
• GitHub, Route and orchestrate models with GitHub Copilot Fusion, 28 October 2026
• OpenAI, API changelog, accessed 5 November 2026
• Anthropic, Pricing, accessed 5 November 2026
• Google, Gemini 4 Carbon: Our next frontier model, 28 October 2026
• Autoriteit Persoonsgegevens, AP informeert incassobureaus over verantwoorde inzet AI (Dutch), 29 October 2026
• Accountant.nl, AP waarschuwt voor risico's bij inzet AI door incassobureaus (Dutch), 30 October 2026
• Microsoft Learn, November 2026 announcements, Partner Center, 3 and 4 November 2026
• Google AI Studio, Changelog, accessed 5 November 2026
• Anthropic, How Claude marks AI-generated content, accessed 5 November 2026
• Court of Justice of the EU, Opinion: Case C-250/25 Like Company, 3 November 2026
• GSK Stockmann, AI Spotlight: Can AI chatbots serve up the news?, 3 November 2026
• OpenAI, ChatGPT Enterprise and Edu release notes, 4 November 2026
• OpenAI, Manage dots permissions and capabilities, accessed 5 November 2026
• OpenAI, Introducing Images 2.6, 4 November 2026
• Mistral, Le Chonk brings the brains. Studio puts it to work., November 2026