One Hacker, $5,000, 100+ Websites: Four Defenses Against AI Agent Cyberattacks
On October 19, Reuters reported a case that rattled the security community: a single hacker handed most of the work to AI agents and, within a few weeks, broke into the websites of more than 100 companies, spending only about US$5,000 on AI along the way. This is not a movie plot. It is a sign that AI agent cyberattacks are now real. For brand owners, the point is not to panic but to take stock: how many doors does your company have open on the internet, and which of them has nobody been watching?
AI Agent Cyberattacks in Practice: One Person, One Prompt, Over 100 Sites
According to Reuters and security researcher Snorre Fagerland of NorForge, the campaign looked roughly like this:
- Scale: Starting on September 25, the AI agents attempted to break into more than 1,200 websites and breached more than 100 companies, mostly e-commerce sites in the US, Canada, Australia and Japan. Victims identified by Reuters include Deere, SharkNinja and Cummins, and two smaller businesses acknowledged being compromised.
- Cost: Fagerland estimated about US$5,000 in AI usage, or roughly US$50 per successful intrusion.
- Effort: The hacker's instructions ran to fewer than 100 Chinese characters. The agents then worked on their own for hours, and successful intrusions took about 23 hours on average.
- Method: In one case, after finding no weakness in a large company's main website, the agents searched related online records, found a public-facing control panel and, after hours of trial and error, exploited a subtle flaw in how it handled data to run commands on the server.
Fagerland put it bluntly: this is not amateur work. What once took a team weeks can now be repeated at scale by one person with AI agents.
The Weak Spot Is Often the Site Next Door: Support Sites, Campaign Pages and Vendor-Run Back Ends
The most telling detail for brands is how the victims responded. SharkNinja said the breach hit a standalone product-support site. Cummins said it was investigating a website run by a third-party vendor. In other words, what gets breached is often not the main homepage but the site next to it.
Most companies have corners like these:
- Campaign or giveaway pages from years ago that were never taken down
- Small sites for dealers or customer support, maintained by an agency whose contract ended long ago
- Staging sites, old admin panels and management consoles whose URLs were never published but can still be found
These corners used to be safe-ish because human attackers have limited time and go after big targets. AI agents never get tired and cost little, so they can work through them one by one. An unmaintained website is now an open invitation.
Four Lines of Defense: Inventory, Update, Lock Down, Monitor
You don't need to buy an expensive security suite overnight. Getting the basics right blocks most automated attacks.
1. Inventory: List Every Site and Admin Panel You Own
Put your main site, campaign pages, online store, support sites, staging sites and anything hosted by vendors into one list, with an owner, a maintainer and a last-updated date. The site you can't list is the most dangerous one. If it's no longer in use, take it down.
2. Update: Patch Systems, Plugins and Packages on a Schedule
Many breaches exploit flaws that were disclosed long ago but never patched. Set a regular update rhythm for your CMS, plugins and code dependencies, and write into vendor contracts who is responsible and how often.
3. Lock Down: Don't Expose Admin Areas to the Whole World
Restrict admin panels, database tools and staging sites by IP where you can, add two-factor authentication, and always change default credentials. Nobody outside your team needs to see the back-end login.
4. Monitor: Make Sure Someone Gets the Alert
AI agent intrusions often take hours of attempts and leave a trail of unusual activity. Set up alerts for failed logins, traffic spikes and modified files, and name a person who will actually receive and act on them.
If part of your website or internal tools was written with AI, see our related piece, It Runs, but Is It Safe? Four Security Questions to Ask Before Commissioning Custom Software.
AI Patrols, People Decide: Make Security Part of Everyday Website Operations
If attackers use AI, defenders can too. Scanning for weaknesses, organizing logs and spotting anomalies are repetitive, time-consuming jobs that suit AI well. Some things, though, must stay with people:
- AI handles: continuous vulnerability scans, consolidating login and traffic logs, flagging anomalies, reminding you which packages need updates
- People handle: deciding which sites to retire, which admin panels to close, whether to pause a service during an incident, and how to explain it to customers
Security is not a one-off project; it is part of running a website. The best time to design defenses in is before launch. The second-best time is now.
At TitApeX, we believe "AI delivers perfection, people make it moving," and security works the same way: let AI patrol around the clock while people keep the final call. If you'd like to take inventory of your company's websites and admin panels, or build security into a new site or system before launch, explore our services or get in touch.
Comments