// this week in AI
This week in AI: ChatGPT ads arrive in Italy

Every week dozens of AI releases come out. 95% of them are no use to you. Here you'll find the 5% that matters, explained without tech jargon — with a single question in mind: what changes for anyone running a business? This week there's a date to mark on the calendar, August 24, and a couple of uncomfortable numbers: what an AI-powered scam costs, and how many people now look at AI with suspicion.
1 · From August 24, ads inside ChatGPT arrive in Italy
Last week we covered the test of ads in ChatGPT being extended to five non-European countries. The update came on August 19 and it's big: OpenAI is bringing advertising into ChatGPT in 31 European countries — Italy, Germany, France, Spain, the Netherlands, Austria and the Nordics — starting August 24. It's the largest expansion of its ad business so far. Only users on the Free and Go plans see the ads: Plus, Pro and Enterprise stay clean. To buy space, at first you go through OpenAI's Ads Solutions team, agencies and technology partners; self-service access from the Ads Manager is announced for later. The tools, though, are all already there: CPM and CPC bidding, conversion optimization, geographic targeting, custom audiences, tracking pixel and Conversions API.
2 · Google changes the engine that summarizes answers, again
On August 13 Google introduced Gemini 3.7 Flash as its workhorse model for code and agents. The next day it announced the model is going into AI Mode in Search. For now the rollout is global but limited to Google AI Pro and Ultra subscribers, in English, and the model is picked manually from the menu in the “Ask anything” bar; free users will get it later. Google claims improvements in following instructions and in understanding searcher intent. The interesting data point is the pace: in May the default AI Mode model had become Gemini 3.5 Flash, and less than three months later there's already another engine under the answers.
3 · One in four breaches is AI-enabled (and costs a million more)
CNBC picked up IBM's Cost of a Data Breach 2026 and the picture is stark: one in four malicious breaches is now assisted by artificial intelligence, 56% more than the previous year. These breaches cost an average of $6 million, about a million above the global average of $4.99 million. The two main ingredients: deepfake impersonation and AI-written malware. The study analyzed 602 organizations between March 2025 and February 2026. There's also a flip side, and it's encouraging: those who use AI and automation in defense save nearly $2 million per breach, but one organization in four hasn't adopted them yet.
4 · The public hasn't warmed to AI: 52% are more worried than excited
TechCrunch lined up the 2026 surveys and the verdict is uncomfortable for the industry. According to Pew Research, 52% of Americans say they're more worried than excited about artificial intelligence, versus 37% in 2021: five years of extraordinary products and trust has gone down, not up. A CNBC survey reports that the majority of 18-to-34-year-olds don't trust the leaders of the AI industry; for Economist/YouGov, over 70% think AI is moving too fast. The article's thesis is that it isn't a communication problem: people see the costs — jobs, works used without permission, data taken without consent — and the benefit in everyday life far less.
5 · ChatGPT opens its version for minors (and starts guessing your age)
Since August 18 OpenAI has been rolling out ChatGPT for Teens worldwide, an experience dedicated to 13-to-17-year-olds with tighter limits on self-harm content and on romantic or sexual conversations. The full rollout takes about two weeks. No documents need uploading: it switches on by itself when the declared age — or OpenAI's age-estimation system — indicates a minor. Parents can link the account and set quiet hours, disable voice, memory and image generation, and opt out of having conversations used for training; reading the chats, however, they cannot. There's also a Study Mode that, when it sniffs out copied homework, brings the kid back to redo the reasoning step by step.
In short
The thread of the week is that AI is ceasing to be a laboratory and becoming a place where things get sold: ads in chat from August 24, an answer engine changing under Google's hood, scams costing millions and a public that trusts it less and less. In the middle of all this, the thing that protects a small business is always the same and can't be bought: getting found with real content, disclosing when AI is involved, and verifying by phone whoever asks you for money.
Frequently asked questions
When are ads coming to ChatGPT in Italy?
From August 24, in 31 European countries, Italy included. Ads are shown only to users on the Free and Go plans; Plus, Pro and Enterprise stay ad-free. At first, ad space is bought through OpenAI's Ads Solutions team, agencies and partners; self-service through Ads Manager will come later. In the meantime, ask ChatGPT what one of your customers would ask.
How can a small business protect itself from AI-powered scams?
One in four malicious breaches is now AI-assisted, and the fake-supplier scam asking you to change the IBAN can arrive in the owner's cloned voice. The best defense is free: verify every change of bank details by calling back the number you already had in your contacts. Then two-factor authentication, a tested backup and up-to-date plugins.
Do you need special optimization to show up in Google's AI answers?
There's no trick to chase: the model that summarizes answers in AI Mode changes every few months, and in August Google announced Gemini 3.7 Flash, after 3.5 Flash in May. The winners are pages that answer real questions in full, with data, proper names and prices, instead of stacking keywords.
A half-hour chat, no strings attached, in plain language.
Let's talk →
Sources: Search Engine Land (ChatGPT ads in 31 European countries) · OpenAI (ChatGPT Ads expands across Europe) · Search Engine Land (Gemini 3.7 Flash in Search) · Search Engine Journal (Gemini 3.7 Flash in AI Mode) · CNBC (breaches and AI) · IBM (Cost of a Data Breach 2026) · TechCrunch (trust in AI) · TechCrunch (ChatGPT for Teens) · Axios (ChatGPT for Teens).