Google’s AI Connections Can Shape Ads, Plus 2 Stories
Google’s latest privacy-policy change doesn’t mean the company has quietly started reading every Gmail inbox and photo library to build ad profiles. That would be the loud, alarming version of this story. The real version is narrower, opt-in, and still worth checking.
The Google data privacy question here is less about whether an AI assistant can find a dinner reservation or remember a trip. It’s about what else becomes possible after you let one service draw context from another. Permission can be voluntary and still deserve a slower look, especially when the product explaining the permission also benefits from your saying yes.
Since October 16, Google’s policy has said it won’t show personalized ads based on content from Drive, Gmail, or Photos unless you choose to make data from those services available to another Google service that supports personalized ads. The obvious route is Personal Intelligence, the feature that can connect Search or Gemini to apps such as Gmail and Photos so the AI can answer questions using your own context.
Google says those connections are off by default, that users choose the exact apps, and that advertisers don’t receive inbox contents, photos, names, or email addresses. Google also told Ars Technica that ads won’t be based directly on an inbox or photo library and that sensitive information won’t be used to tailor ads. But Google’s own Photos privacy hub gives a concrete example: if you enable Personal Intelligence in Search and connect Photos, Search may process some of the data you permit it to access to personalize content, including ads.
That distinction matters. The system doesn’t need to hand a photograph to an advertiser for the photograph’s meaning to influence what the service shows you. A summary, an inference, or the context around an AI answer can still affect personalization. The useful AI feature and the ad system aren’t identical, but the data boundary between them is no longer the simple wall many people assumed it was.
I don’t think the right response is to panic and abandon every Google product. I do think “connect everything so the assistant knows me” is a much bigger permission than the friendly setup screen makes it sound. Open Gemini’s settings, look under Personal Intelligence and Connected Apps, and leave connected only the services whose benefit is worth the access. If you never enabled the feature, there’s no emergency here. The useful action is verification, not a digital bonfire.
Turning on AI is the easy part
Deciding what an AI tool should be allowed to see, who should use it, what work it should perform, and what happens when it gets something wrong is the more interesting problem. Raymond Tec helps businesses connect and automate the tools they actually use without treating every new feature like a button that obviously needs to be switched on.
Two exploited flaws need precise patching
CISA added two very different vulnerabilities to its Known Exploited Vulnerabilities catalog on October 23. Both deserve quick attention, but neither justifies warning every Nextcloud or VeloCloud user indiscriminately.
The first, CVE-2026-110447, affects the Collabora Online Built-in CODE Server app for Nextcloud, package name richdocumentscode, before version 26.4.303. An operating-system command injection flaw in proxy.php can let an attacker run commands on the server when the built-in CODE service isn’t running or has recently restarted. Reports in the Nextcloud community include web shells found on exposed systems. A standalone Collabora Online server, a hosted Collabora service, or Nextcloud without the built-in CODE app isn’t the same exposure.
If that app is installed, update it to 26.4.303 or later now. Before changing anything, take a recoverable snapshot or confirm you have a tested backup that the affected server can’t erase; an attacker with command execution may have more reach than the application itself. Because exploitation has been observed, patching closes the door but doesn’t prove nobody already came through it. Review the Nextcloud integrity scan, web-server logs, unexpected PHP files, and credentials the web-server account could read. If the app isn’t installed, don’t turn a product name into a false alarm.
The second, CVE-2026-114625, affects on-premises Arista VeloCloud Orchestrator. Arista says an authenticated user, including an Enterprise Read Only user, can exploit SQL injection to gain operator privileges and masquerade as a local superuser. That turns a stolen low-privilege portal account into a path toward control of the orchestrator and its managed data.
Hosted and dedicated VeloCloud services were affected but have already been patched by Arista. On-premises operators need at least 5.2.3.18, 6.1.3.10, or 6.4.2.12 in the corresponding release branch. Arista says the flaw is actively exploited. Check the advisory’s indicators and logs, restrict management access, and patch the exact affected system. Specificity isn’t a footnote here. It’s the difference between useful security work and sending everyone on a scavenger hunt.
Business IT goes well beyond the website
Your business also depends on workstations, cloud accounts, browsers, Wi-Fi, remote access, collaboration tools, and all the other technology that quietly becomes infrastructure. Raymond Tec works across that whole stack, whether the problem lives on a server, on a desk, or somewhere in between.
Meta buys three weeks on disputed business terms
Meta has delayed its updated commercial terms from October 30 to November 20 after objections from publishers, advertisers, and U.S. Senator Marsha Blackburn. Critics say the language could give Meta broad authority over content connected to Facebook and Instagram business accounts, including external websites. Meta disputes that reading. It says the update applies to advertising and business products, doesn’t change its web-crawling practices, and doesn’t override a site owner’s robots.txt preferences.
The practical issue isn’t only copyright language. It’s control: a small business may have given an agency, former employee, or platform integration access to both the social account and the website. This is a good reason to identify the actual owner of each account and the person responsible for accepting terms.
The delay is a material change to the terms dispute covered here last week, but it isn’t a resolution. If your business relies on Meta products, use the extra time to read the version that will actually take effect, confirm who controls your website’s crawler rules, and decide what content you’re willing to connect to those accounts. I wouldn’t make a technical change based on the accusation alone, and I wouldn’t treat a postponed deadline as the concern disappearing either.
Still in a reading mood? The Raymond Tec News archive covers security, AI, small-business technology, policy, and the places technology collides with ordinary life — without requiring a computer-science degree to get through it.
Sources / Further Reading
- Google Privacy Policy
- Google Photos: Gemini features in Photos privacy hub
- Gemini release notes: Personal Intelligence connections
- Ars Technica: Google’s privacy-policy clarification
- The Verge: Personal Intelligence and personalized ads
- CISA Known Exploited Vulnerabilities Catalog
- CVE-2026-110447 record
- Nextcloud community warning for Built-in CODE Server
- Arista Security Advisory 0199
- Reuters: Meta delays updated commercial terms
Photo by henry perks on Unsplash.
