mekat.
Try a lesson

The internet had a rough week. Here’s your plan

This one couldn’t wait for Monday, so here’s an early edition. When you use AI, or any app, you’re trusting more than the words on your screen. You’re also trusting the software behind it and the access you’ve given it. On Wednesday, October 21, three stories tested that trust at once: a serious flaw in the software behind a huge share of the web, a popular AI tool that briefly shipped with a password stealer inside, and a hosting company warning that customers’ secret keys may have been exposed. None of this means every AI user was hacked, and for most people there is nothing to install. But the fallout tends to land in your inbox. Here’s the good, the bad and the ugly. Last checked: October 21, 2026, about 7:30 p.m. Arizona time.Mekat Monday By the Mekat team

The Good

The headline flaw has a nickname, PHP2Shell. PHP is the programming language running behind the scenes of many websites, including a lot of WordPress sites. On October 21 the PHP project released fixed versions and an advisory saying the two flaws can let an attacker run their own code on a web server without logging in. That’s about as serious as web bugs get.

The hopeful part is how they were found. Google Cloud says Wiz Research found the flaws with help from several AI models, and that Google’s Gemini helped with finding them, detecting attacks and checking the fix, according to Newsweek. So this week AI was also on the defenders’ side: helping the people who find holes before criminals do. The fixes themselves came from the PHP project, which released them the same day.

Big hosts moved fast, too. Automattic said, according to Newsweek, that sites on WordPress.com, WordPress VIP, WP Cloud and Pressable were already protected, so those customers don’t need to do anything.

What this means for you:

  • If you just visit websites, there is no PHP2Shell update for your phone or laptop. The flaw lives on servers, not on your devices.
  • If you run a website, ask your host or developer one question: “Is our PHP updated for PHP2Shell?” Automattic’s point matters here: updating WordPress, themes or plugins alone doesn’t fix it. The PHP version underneath has to be patched.

The Bad

When one piece of shared plumbing breaks, a lot of the internet feels it at once. The PHP project says every PHP installation that serves web requests is at risk until it’s updated. Some site owners chose to go offline rather than wait. Newsweek reports that the Wikimedia Foundation took several systems offline as a precaution, and, citing Reuters, that government websites in the Netherlands and Finland went offline. A security researcher, Kirill Firsov, had already posted a php2shell video on October 20 with the caption “Tomorrow is a helluva day.” Understatement of the week.

One number gets passed around: Wiz estimates PHP powers about 70% of websites. That’s how common PHP is, not how many sites were hacked. Newsweek makes the same point.

How to handle it:

  • Expect some “down for maintenance” pages this week. That’s a site being careful, not a sign that you were hacked.
  • Don’t install “PHP2Shell protection” apps or browser add-ons. Nothing like that exists for regular users, so anything selling one is a scam.

The Ugly

The other two stories share a target: the keys that let apps talk to AI. Think of them as the passwords that apps use, not people.

First, LiteLLM, a popular open-source tool that many apps use to connect to AI models. LiteLLM says version 1.107.0, published at 11:39 UTC on October 21, contained malware that steals credentials and can load more code. Security firm Chainguard notes it ran as soon as Python started, so simply installing it was enough to set it off. LiteLLM says the bad version was pulled from the main download site within about two hours, and that it was the project’s second incident this year after one in March. Pulling it doesn’t undo the damage on machines that already ran it.

Second, Railway, a company that hosts apps for developers. Railway says it detected unusual activity in the systems that store customers’ secret settings, and that an attacker may have reached some of them in readable form. It paused all deployments, told customers to replace every key stored there, and said it believes the attack was part of a series aimed at platform companies, with the goal of getting at AI providers’ keys. Railway’s status page still showed builds paused late on October 21.

Nobody is saying these were one coordinated attack. But Railway’s own explanation points to the pattern: attackers want the keys that unlock AI services. If an app you use was affected, it’s that company’s job to replace its keys. Your job is to not get phished in the aftermath. Breach news is a gift to scammers. Expect emails that say “Your account was affected, reset your password here.” The FTC’s advice holds: when a message asks you to click a link or open an attachment, stop and ask whether you even have an account there, and it lists claims of a problem with your account among the common phishing stories.

If you want the full list of everything else that broke this week, one developer’s running roundup on X has it, though not every item there has been confirmed by the companies involved.

How to protect yourself:

  • Use a different password for every account. A password manager makes that painless.
  • Turn on two-step sign-in (MFA) for email, banking and anything with a card attached.
  • If you get an unexpected “reset your password” email, don’t click it. Open the app or type the site’s address yourself, and check from there.
  • Keep passwords, recovery codes, one-time codes and card numbers out of AI chats, and out of the screenshots you upload. A chatbot is a terrible place to keep a secret, and you can’t see which tools sit behind it.

Scout’s take

A meerkat on lookout duty doesn’t panic at every shadow. It keeps watching and knows exactly where the burrow is. You don’t need to become a security engineer either. There’s nothing to install this week, so do the boring stuff that actually works:

  1. Turn on two-step sign-in for your main email account. It’s the key to every other account.
  2. Change any password you’ve reused on more than one site, starting with the important ones.
  3. Give apps and AI tools only the access they need. Look at your connected accounts and remove anything you no longer use.
  4. If a service reports an incident, open its official website yourself and follow its guidance there. Never trust a surprise “security fix” link in a message.

The lesson isn’t “stop using AI.” It’s “understand what you’re trusting.”

Want the habits to stick? Mekat’s lessons on phishing, voice clones and deepfakes, and on hidden instructions and account safety, give you practice with realistic examples. Try a free mini lesson on mekat.ai or see the full learning path. See you next Monday.

Sources

Published by Mekat, Martinez Solutions LLC, Glendale, Arizona. Every factual claim is checked against the sources above before publishing. Spotted a mistake? support@mekat.ai