For five years Google promised it would never use what's inside your Gmail, Drive or Photos to pick the ads you see. On October 16 it added one condition to that promise. The condition is its own AI.
It's a quiet Monday otherwise, so today is short and pointed: Google's new ad exception, a $15 billion bet on the cheapest lab in AI, and a countdown on Meta's business terms that ends this Friday. Let's go.
Google's ad promise now has an AI-shaped exception
Since 2021, Google's privacy policy has said plainly that content from Gmail, Drive and Photos is not used for personalized ads. The new version, in effect since October 16 and first spotted by Android Authority, keeps that line but adds a door: the promise holds unless you choose to share that data with other Google services that support personalized ads.
The door has a name. Personal Intelligence, which Google launched in Gemini in January, lets Gemini and AI Mode in Search read your connected apps so the answers fit your life. Connect Gmail to get a smarter answer about your flight, and that connected content can now help choose your ads too.
Google's side of it: nothing is used for ad targeting by default, and advertisers never see the contents of anyone's email, files or photos. To close the door, go to your Google Account, open Data & privacy, then Connected content, and disconnect the apps. Android Authority also noted that older data can sit in your Search history after you disconnect, so clear that too if it bothers you.
Why should a seller care? Two reasons. First, your own inbox. Supplier invoices, customer addresses, order confirmations: if you connected your business Gmail to Gemini for convenience, check that setting today. Second, the ad side. Every big platform is moving toward ads chosen from what people say to their AI, not just what they click. We saw it when ChatGPT ads started matching your offer to the conversation, and now Google has written the same direction into its policy. Your ad money will buy sharper targeting. You just won't see the inputs.
DeepSeek raised about $15 billion, days after a US ban
Bloomberg reports that DeepSeek raised roughly 100 billion yuan, about $15 billion, with Tencent and battery maker CATL among the biggest backers. Reuters counted more than 15 investors, including Meituan, Midea and Geely, and said the target grew twice along the way, from 50 billion yuan to 80 billion and then to 100 billion. Reuters also put the value before the new money at about 500 billion yuan, and said some details could still change before the round fully closes.
The timing is the story. On October 20, the US Department of Homeland Security published an order telling federal civilian agencies to remove DeepSeek products from their systems, including third-party tools built on its models. The order covers government systems only. Companies and the public can still use DeepSeek. Two days later, investors handed it double its own target.
So what does a shop owner take from a fight between governments and billionaires? Mostly good news, with one lesson attached.
The good news is price. DeepSeek is the lab that keeps dragging the cost of AI down, and $15 billion means it can keep doing that. Cheaper models end up as cheaper tools for you, which is why the numbers in our customer support automation cost breakdown keep moving in your favor.
The lesson: the model under a tool can turn into a political problem overnight. If you rely on any AI tool for selling, ask the people behind it one plain question. If the model we use today disappears tomorrow, can you switch without breaking my setup? A good answer is short and boring.
Four days until Meta's new business terms apply
On October 30, Meta's updated Commercial Terms take effect for anyone who uses Facebook, Instagram, Messenger or WhatsApp for business. The headline clause lets Meta crawl websites associated with your business presence on its apps.
A week ago we wrote that Meta backed off its AI-training clause, but only halfway. That picture is still blurry. Search Engine Journal reported that the word "train" was dropped from the list of uses. PPC Land pushed back, saying the reporting didn't quote any revised text, and it found the training language still there when it reviewed Meta's preview page on October 18. Meanwhile the European Publishers Council has asked the European Commission to look at the terms.
On robots.txt, the file that tells crawlers what they may read, Meta told Nieman Lab it doesn't intend to use the terms as a legal basis to ignore it. Meta runs separate crawlers for ads and for AI, but its AI crawler uses a single token for both training and indexing, so blocking it isn't a clean yes or no.
Before Friday, spend fifteen minutes on this:
- Read the terms in your business account's preview page yourself. Don't rely on headlines, including ours.
- Check what your website actually shows. Wholesale prices, supplier names, customer photos you never meant to publish: if it's public, assume a crawler will read it.
- Decide on purpose about Meta's AI crawler in your robots.txt. Blocking it may also affect how Meta indexes your site, so weigh both sides.
- Keep a copy of the page as it stands on the 30th, in case the wording shifts again.
One small one: your Gemini Gems are moving
If you built Gems in Gemini for product descriptions or caption ideas, note the date. On November 17, Gems move into the Gemini app's Settings panel, and skills become the main way to save reusable instructions. You can still use Gems after that. Skills are rolling out in the Gemini app through mid-November, while the Workspace version is delayed until further notice, and Google says app skills won't carry over to Workspace automatically. My advice: copy your best Gem instructions into a plain note now, so a menu shuffle can't cost you a working prompt.
What this week means for your shop
Every story today asks the same thing: who gets to read your business, and on whose terms? Google wants your inbox for better answers and better ads. Meta wants your website. Governments are deciding which models are allowed where. You can't stop the platforms from asking, but you can decide what you hand over, on purpose. That's the idea behind Vardast: an AI assistant that answers your customers on Instagram, WhatsApp, Telegram and your website, using only what you give it, with you setting the rules.