Daily AI brief · October 20, 2026
One Hacker's AI Agents Broke Into 100+ Websites
One person used AI agents to break into more than 100 company websites, mostly online stores, for an estimated $50 a break-in, and Google is now testing ads in the AI Mode that reads your connected apps.
One hacker's AI agents broke into more than 100 company websites, most of them online stores (Reuters, Cybersecurity News, Gizmodo) Reuters reported on Monday that a Chinese-speaking hacker pointed AI agents at more than 1,200 websites starting September 25 and got into more than 100 of them, mostly online stores in the US, Canada, Australia and Japan. The researcher who found it, Snorre Fagerland of NorForge, spotted the attacker's control screens left open on the internet. The instructions were short: find a way to run code on each site without signing in, then plant a hidden file uploader. The agents then worked on their own for about 23 hours per successful break-in, hunting for unsafe file uploads, login bypasses and old software parts. Deere, SharkNinja and Cummins were named among the targets; SharkNinja said only a separate product-support site was hit and it was contained. NorForge estimates the whole run cost about $5,000 in AI fees, roughly $50 per break-in, though it says those numbers are rough, and Reuters couldn't tell how much data was taken. For a small business, the price is the news: at $50 a try, nobody needs to pick you on purpose anymore, and a store or booking site running on software that someone installed years ago and never updated is exactly what these agents look for. Stores on hosted platforms that update themselves are in a better spot than sites running on your own server.
Google is testing ads in AI Mode for people who connect their apps (Search Engine Land, Search Engine Roundtable) Google has started testing ads in AI Mode, its chat-style search, for people who turn on Personal Intelligence, the setting that lets it use Gmail, Photos and other connected apps. In March Google said that version would stay ad-free. Google says ads won't be based directly on someone's inbox or photos, but summaries and conclusions drawn from those apps may now be used to pick ads, advertisers get no access to the apps, and sensitive topics like health and religion aren't used. Google hasn't said which countries, which ad formats, or whether advertisers get any setting or report for this traffic. If you buy Google ads, the practical part is that your ads may start showing up inside AI answers you can't see in your reports yet. If you don't, the people searching for what you sell are still more likely to see a paid competitor above any free mention of you.
For whoever looks after the website (forward this): send your web person or hosting company three questions today. When was the site's software, including every plugin or add-on, last updated? Which add-ons are installed that nobody uses anymore, and can they come off? Is the admin login page protected with two-step sign-in? If the answer to the first one is "not sure," that is your answer, and the update goes on this week's list.
If you run a home services company: follow one call from start to finish. A homeowner's furnace quits, they ask Google's AI what to do, see an ad, and call. Whoever answers books the visit and asks one question: "Where did you find us?" Put the answer in a column next to the job. An AI tool can tally that column every Friday, but the owner decides what the ad money does, and until Google gives you a report for AI Mode, that column is the only one you'll have.
Don't move ad money because of a headline. Google hasn't said where these ads run, what they look like or how they're counted. Keep your budget where it is, watch your cost per call for a month, and decide with your own numbers.
Where it's going: AI is making it cheap both to attack a small business and to sell to its customers, so the businesses that keep their website software current and track where every call comes from will spend less and lose less than the ones guessing.
- AI agents broke into 100+ websites: [Reuters via Superpower Daily, Oct 19](https://superpowerdaily.com/posts/hacker-used-ai-to-breach-sites-at-more-than-100-companies-reuters-finds) · [Cybersecurity News, Oct 2026](https://cybersecuritynews.com/chinese-hacker-use-ai-agent-to-hack/) · [Gizmodo](https://gizmodo.com/the-fear-of-a-one-man-band-weaponizing-ai-agents-is-becoming-reality-2000827754) · [WinBuzzer, Oct 20](https://winbuzzer.com/2026/10/20/hacker-used-ai-agents-to-breach-100-plus-companies-websites-xcxwbn/)
- Ads in Google AI Mode with Personal Intelligence: [Search Engine Land, Oct 16](https://searchengineland.com/google-begins-testing-ads-in-ai-mode-for-users-with-personal-intelligence-enabled-496990) · [Search Engine Roundtable, Oct 19](https://www.seroundtable.com/google-ads-ai-mode-personal-intelligence-42320.html)