So your business doesn’t break while you’re busy running it.
A 5-minute weekly brief that tells you what to ignore, what to fix, and what can wait.

This week’s issue is about the accounts and tools that keep your business visible: your Facebook and Instagram business pages, your AI assistants, your phone, your website, and more.

The thread running through all of it is simple: attackers are not always breaking in anymore. Sometimes they use a real Google address, a connected AI plug-in, a fake support call, or location data your phone shared in the background.

Let’s make this practical. 👇

If you…

Use Facebook, Instagram, Meta Business Suite, or paid ads for your business. Take ten minutes today to check your account security.

Should you care?

✅ YES – urgently – You use Facebook or Instagram to reach customers, run ads, manage a community, receive messages, or maintain reviews. Losing the account could interrupt sales and customer trust.

🤷‍♀ MAYBE – worth checking – You have dormant pages, old ad accounts, or former employees who may still have access. Those old accounts can still be abused.

❌ NO – low priority (for now) – You have no Meta business presence and no old business pages tied to your name, brand, or email address.

What’s happening (plain English)?

Security researchers reported that a phishing campaign stole roughly 30,000 Facebook accounts by sending fake Meta support emails from a legitimate Google address. The emails warned that business pages would be permanently deleted, then walked people through fake CAPTCHAs, fake Meta privacy pages, and forms that collected passwords, two-factor codes, dates of birth, and even government ID images.

The scam works because the email looks trustworthy. It comes from a real Google-owned sending address, not a strange throwaway domain, and it uses a fear-based message: appeal now or your page disappears. For a small business owner, that pressure is powerful. Your page may hold years of photos, reviews, messages, ad audiences, customer questions, and proof that your business is real.

The fake pages also use familiar tools like Google Drive, Netlify, Vercel, and Canva, which makes the scam feel more legitimate. That matters because “just check the sender” is no longer enough when attackers are abusing real platforms to carry the attack.

Do this now

Researchers are finding serious security weaknesses in the plug-ins and connectors that let AI assistants touch your email, files, calendars, code, CRM, and business apps. Help Net Security reported on Noma research finding that 1 in 4 MCP servers opened AI agents to code execution risk, and OX Security previously reported remote-code-execution vulnerabilities across tools including Flowise, LiteLLM, Agent Zero, Windsurf, and others.

In plain English: the risk is not just that a bad person sends you a phishing email. The risk is that your AI reads something booby-trapped, such as a document, lead form, pull request, CRM record, or webpage, and silently follows hidden instructions inside it.

This is especially relevant if you are experimenting with AI agents, automation tools, browser assistants, coding assistants, or “connect your Gmail/Drive/CRM” style features. The more permissions the assistant has, the more damage a bad instruction can do.

Why it matters

For a small business, AI tools are attractive because they save time. They can draft emails, summarize meetings, update records, search files, and move work between systems. But if an AI tool has “allow all” access, a compromised connector can become a side door into the same systems you are trying to protect.

Do this now

This is more common than people think. Website hacks are often designed to be invisible to the owner. You may see your normal homepage, while customers or search engines see spam pages, pharmacy links, gambling pages, malware redirects, or fake checkout pages.

What usually causes it?

For WordPress sites, the usual culprit is an outdated plugin, theme, weak admin password, unused admin account, or old hosting control panel. For Shopify and other hosted platforms, the core platform is handled for you, but third-party apps, themes, admin accounts, and staff permissions are on you.

The simple prevention plan

If your site is already hacked, do not just delete the weird pages and move on. Attackers often leave backdoors. Take the site offline if needed, call your host’s emergency support, restore from a clean backup, rotate passwords, and review every admin account.

Bottom line: your website is part of your storefront. Treat it like a cash register, not a brochure.

Microsoft says phishing is bypassing normal MFA

Microsoft documented a large phishing wave that targeted more than 35,000 users at 13,000 organizations with polished “code of conduct” lures and phishing pages designed to steal both passwords and post-MFA session tokens.

Fix: Use passkeys or hardware security keys for admin, finance, email, and payroll accounts when possible. Train your team that HR, compliance, payroll, and legal threats should be verified through a known phone number or internal channel, not through a link in the email.

Instagram encrypted DMs are going away

Meta is removing Instagram’s optional end-to-end encrypted direct messages instead of a mode where only the people in the conversation can read the message content. The company says very few people used the optional feature and is pointing users who want encrypted chats to WhatsApp instead.

Fix: Do not use Instagram DMs for sensitive business conversations, payment details, legal issues, customer complaints, private images, or anything you would not want preserved in a platform-accessible message stream. Move sensitive conversations to a tool with default end-to-end encryption and clear retention controls.

Fake Claude AI sites are pushing malware

Hackread reported on Sophos X-Ops research about a fake Claude AI site using the domain claude-pro to trick people into downloading a fake “Claude-Pro Relay” tool. The download installs Beagle malware, which can run commands, upload and download files, and browse directories on the infected computer.

Fix: Download AI tools only from official vendor sites, not search ads, sponsored links, or “pro” download pages you find through Google. If someone on your team installed a Claude desktop helper from anywhere other than Anthropic’s official site, treat that machine as suspicious and scan it.

Shared-hosting control panels are being targeted

Did you update cPanel per last week’s newsletter? Time to do it again. The Hacker News reported that three more critical cPanel/WHM vulnerabilities, with compromised servers used for scanning, botnet activity, and ransomware deployment.

Fix: If you use shared hosting, ask your host if cPanel/WHM is patched. Also turn on 2FA for your hosting and control-panel logins.

Fake IT-support calls are stealing SaaS accounts

CrowdStrike research covered by The Hacker News described extortion groups calling employees, impersonating IT support, and directing them to fake SSO login pages to capture credentials and one-time codes.

Fix: Tell your team: Noone will ever cold-call and ask you to log in through a link. Hang up, then call back using a known number if you think it may be legit.

Vibe-coded apps are leaking private data

WIRED reported that researchers found more than 5,000 AI-built apps on the open web with little or no authentication, and nearly 2,000 exposed private information such as hospital staffing details, customer chatbot logs, sales documents, financial files, shipping records, and go-to-market materials. The pattern is especially risky for nontechnical founders using AI app builders because a quick prototype can quietly become a public data leak.

Fix: Before sharing any AI-built app, open it in an incognito window and confirm what a stranger can see. Add authentication, remove real customer or company data from test apps, and keep prototypes private until someone has reviewed access controls.

The FTC settled a lawsuit against data broker Kochava after alleging the company sold precise phone-location data that could reveal visits to sensitive places such as mental health clinics, reproductive health providers, places of worship, and domestic violence shelters. Your location patterns can reveal more than you intended.

Do this today

The goal is (usually) not to disappear from the internet. The goal is to stop giving away more than you need to or intend to.

This week’s theme is permissions. Who can access your Meta account? What can your AI tools touch? Who still has website admin rights? Which apps know exactly where your phone goes?

None of these checks are glamorous, but they are the kind of small, practical actions that prevent expensive chaos later.

Reply and tell me. I read every response.

~Alexia

P.S. If one of these checks raises a question, bring it to office hours — free and private using, Fridays at 1pm ET. Sometimes a quick 8-minute conversation is enough to figure out what matters and what can wait. Add office hours to your calendar and drop in when you have a question. (Some folks have asked whether ro.am drops you right in with me, face to face. No, don’t worry, you enter a lobby and I let you in. No surprises for either of us. 🙂)