AI in October 2026: what changed and what matters if you work in Europe

Carlos Imlau, AI Product Builder · Certified trainer (CCP) · AI applied to work · 2026-11-03

What changed in AI in October 2026? For people at work, three things: AI moved deeper into everyday tools (Claude started editing Google documents and spreadsheets), free tiers got thinner (Gemini without a plan now mostly uses a smaller model) and European rules started to bite, with the first AI Act enforcement and watermarks on generated text. In Portugal, where I live and work, the month brought numbers that confirm what you see on the ground: many people already use AI, few use it well.

This is the monthly roundup I write for people who don't have time to follow everything. I picked six stories from October and, for each one, I explain what happened, why it matters to a business or to someone just starting out, and where to read the source.

1. Claude now works inside Google Docs, Sheets and Slides

On 6 October 2026, Anthropic launched an add-on that opens Claude in a sidebar next to your file in Google Docs, Sheets and Slides. It is in public beta for all paid plans and installs from the Google Workspace Marketplace, as Anthropic's help page and 9to5Google explain. In Docs it rewrites sections without breaking formatting, in Sheets it writes formulas and builds tables, in Slides it creates slides from your existing theme. There is a mode that previews changes before applying them.

Why it matters: no more copying text into a chat and pasting it back. For a team that already lives in Google Workspace, it is the simplest way to put AI into real work. My advice: turn on the preview mode and review what changes, especially in spreadsheets with numbers that go to clients.

2. Free Gemini became more limited

Since 9 October 2026, people using Gemini without a paid plan get the automatic mode by default: most requests go to the Flash-Lite model and only some harder ones move up to larger models, according to Google's help page. All plans gained an effort setting (low, medium or high); high effort gives more thorough answers but uses more of your limit. Digital Trends sums up the differences: the Pro model is reserved for AI Pro and above.

Why it matters: if your team uses free accounts for serious work, answer quality may have changed without anyone noticing. It is a good moment to decide, carefully, which paid tool makes sense and who needs it. I have written about how to evaluate an AI tool without falling for launch hype.

3. OpenAI began watermarking ChatGPT text in the EU

On 5 October 2026, OpenAI announced it would add an invisible watermark to text generated by ChatGPT and Codex in the European Union over the following weeks, PYMNTS reports. In the API, the watermark is available for some models but off by default. For now, the detector is open only to approved researchers and organisations.

The reason is Article 50 of the AI Act, which requires providers of generative AI to mark generated content in a machine-readable way. Systems already on the market before 2 August 2026 have a grace period until 2 December 2026.

Why it matters: the marking obligation falls on whoever provides the system, not on someone writing an email with ChatGPT's help. But if your company has a product or an automation that generates text for customers, it is worth knowing where that text comes from and whether your provider already complies. I covered the chatbot disclosure rule in my guide to AI agents in customer support.

4. Europe started enforcing the AI Act

On 14 October 2026, Commissioner Henna Virkkunen announced that the AI Office had started enforcing the AI Act: it sent OpenAI and Anthropic formal requests for information on security, external evaluation and how their models are monitored, after the Commission found the documents both companies had provided insufficient, according to Euronews. The case involves two summer incidents with agents during testing. Some legal experts dispute that the law applies to events that happened before the models reached the market.

Why it matters: for a small business, this does not change obligations overnight. It changes the tone. The AI Act is no longer just a document in Brussels; someone is now asking questions. The obligation that already reaches almost every company in the EU is AI literacy, which I explained step by step in my article on AI literacy and the AI Act.

5. Portugal: half of companies use AI, few integrate it

A study by Strand Partners for AWS, released on 19 October 2026 and reported by Jornal Económico, says 51% of Portuguese companies have adopted AI tools and about 80% of those report productivity gains. But only 16% of companies using AI have reached advanced integration, and 58% stay with basic uses.

Three days later, on 22 October, Portugal's SME agency IAPMEI presented PME Digital+: 2.47 million euros, 85% co-funded by the EU's ERDF, for about 600 SMEs and more than 5,000 participants in the North, Centre and Alentejo regions over 24 months, with 48 awareness and training actions on AI, automation, cybersecurity and digitalisation, according to Jornal Económico. It funds training and diagnosis, not investment.

Why it matters: the problem is no longer "starting to use AI", it is moving from scattered chats to processes that work every day. That step is made with the company's real cases, not with yet another demo. It is what I offer in my training for companies in Portugal: working on your team's actual tasks.

6. AI training for 100,000 workers and the IMF's warning

During the debate on Portugal's 2027 State Budget, on 28 October 2026, Economy Minister Manuel Castro Almeida announced AI training for about 100,000 workers, according to Lusa, via DNoticias. No budget or timeline was given.

Earlier, at the IMF and World Bank annual meetings (12 to 18 October), IMF Managing Director Kristalina Georgieva said AI will affect about 40% of jobs worldwide and about 60% in advanced economies, calling it "a slow tsunami", RTP reports.

Why it matters: "affect" is not the same as "eliminate": it includes jobs that AI makes stronger. The difference between the two often comes down to who learns to use the tools with good judgement. If you are just starting, the best answer to a warning like this is not panic, it is a habit: learn a little every week, with real cases.

What to watch in November?

• The 2 December 2026 deadline for generative AI systems already on the market to mark generated content. If you have a product that generates text or images, check with your provider.
• The details of the training for 100,000 workers in Portugal: who pays, who can take part and when it starts.
• The first PME Digital+ actions by IAPMEI, for businesses in the North, Centre or Alentejo.
• Free-tier limits, which keep changing. Before paying for a tool, test it on your own tasks.

Frequently asked questions

Do I need to watermark the texts I write with ChatGPT?

No. The AI Act's Article 50 obligation to mark content in a machine-readable way falls on providers of generative AI systems. Other transparency rules, such as telling people they are talking to a chatbot, may apply to the company that deploys that chatbot.

Is it still worth using free Gemini at work?

For simple tasks, it may be enough. Since 9 October, most requests without a plan go to the smaller model, so test the tasks you use most and compare. If quality dropped, a paid plan for daily users is usually cheaper than the time lost fixing answers.

Does PME Digital+ give money to buy AI tools?

No. Based on what was announced, it funds training, diagnosis and demonstration actions, with free resources on IAPMEI's website. It is not an investment grant.

How do I keep up with this news without losing hours?

Pick one source that filters for you and only test what touches your work. Every Friday I send an email with practical AI tips; you can sign up on the tips page.