In September, Google's Threat Intelligence Group released its Q3 2026 AI Threat Tracker. The headline case: a financially motivated attacker broke into one organization's cloud setup, then used an AI coding assistant, a prompt and a set of agent instructions to plan, build and run a mass credential-theft campaign in under six hours. Thousands of third-party logins were compromised.
Google was also clear about what it has not seen: fully autonomous, hands-off hacking at scale isn't happening in the wild yet. So this isn't a reason to panic. It's a reason to tighten up, because the window between "someone found a weak spot" and "the damage is done" just got a lot shorter.
If you run a law firm, med spa, medical practice or home service company, your website is probably doing more than it did five years ago. It books appointments, captures intake details, connects to your CRM, and may run an AI chat widget or AI front desk. Every one of those connections is a door. Here's how to check yours.
1. Know what your chat widget can reach
Ask your vendor (or us) exactly what the chatbot or AI front desk is connected to: your calendar, CRM, email, intake forms, payment tools. It should only reach what it needs to do its job. A chat widget that books consultations doesn't need access to your whole client database.
2. Find out where the keys live
AI tools talk to each other using API keys, which work like passwords. Google's report specifically calls out stolen AI service keys and agent configuration files as targets. Those keys should never sit in your website's public code, and they should be changed when a vendor, agency or employee leaves.
3. Clean up who has logins
Make a short list of every account that touches your marketing: website admin, hosting, domain/DNS, Google Business Profile, CRM, ad accounts, chat platform. Remove former employees and past agencies. Turn on two-factor authentication everywhere it's offered.
4. Don't collect what you don't need
If your chat or intake form asks for case details, medical history or payment information, ask whether it really has to. Less sensitive data collected means less to lose. Medical and legal practices should confirm with their vendor how that information is stored and who can see it.
5. Keep the basics current
Outdated plugins, themes and forms are still the easiest way in. Confirm updates are happening and that you have recent backups you could restore from.
6. Watch for warning signs
A sudden spike in junk form fills, a new admin user you don't recognize, a chat tool saying things it shouldn't, or an AI usage bill that jumps for no reason are all signals to check now, not next month.
The bottom line
AI is making attackers faster. It's also making good businesses more responsive, and an AI front desk that answers leads at 9 PM is worth having. The goal isn't to avoid these tools. It's to set them up so they only touch what they should, and so you'd know quickly if something went wrong.
Not sure what your website and chat tools are connected to? Start with our free Growth Audit checklist, see our website security, recovery and rebuild service, or learn how our AI Front Desk is set up to capture leads safely.
Source: Google Threat Intelligence Group, AI Threat Tracker Q3 2026, as reported by Help Net Security (September 2026).


