November 19, 2026
One Text Box, 178 Organizations: How a Trusted Script Turned the BBC’s Homepage Into an Attack
Nobody broke into the BBC, Zoom, or CNN. A small tool they all trusted did the damage instead.

By Naro
9 min read
On Monday morning, millions of people opened the BBC website. The address was correct, the padlock was there, and nothing looked wrong. But their browsers were also running code written by an attacker.
The same thing happened on Zoom, CNN, Fox News, NPR, HSBC, Equifax, Salesforce, GitLab, IKEA, KLM, and the crypto wallet maker Ledger. Security researchers counted 178 organizations.
Here is the strange part: almost none of these companies were hacked. Nobody broke into the BBC's servers, and nobody touched Zoom's systems. The attack came through a small tool most people have never heard of.
I want to explain what happened in plain language, what the malicious code tried to steal, and what the incident says about how the modern web is built.
Is the headline true?
The headline claims that one cyber attack hit 178 major organizations. It's true, but not in the way most people picture it.
The number comes from Scott Helme, founder of Report URI, a company that monitors scripts loading on websites. His team collected 1,611 copies of one script from one vendor, Optimizely. Of those, 188 files carried hidden malicious code, and they belonged to 178 organizations. Report URI believes the real number is likely higher, because it could only check the files it found in time.
One fair warning: Report URI sells tools that detect this kind of attack, so it has a commercial interest. Where I can, I compare its numbers with other researchers.
It also matters what "hit" means here:
- Visitors to these sites were served harmful code inside a file the site itself requested.
- It does not mean attackers got into the companies' internal networks. Equifax, for example, said it found no evidence its systems were compromised.
- It does not mean every page of every company was affected. For IKEA, only a single home design tool was listed. Hudson Rock, another security firm, warns that one affected subdomain doesn't prove exposure across a whole organization.
Think of it as one poisoned package delivered to 178 front doors. The houses weren't broken into. The delivery was tampered with.
What is Optimizely, and why did everyone trust it?
If you've ever noticed a shop's button was green one day and orange the next, you've probably seen an A/B test. The company shows half its visitors one version and half another, then keeps whichever sells more. Optimizely is one of the biggest tools for this. Its website says more than 2 million experiments have run on it.
Here is how it works:
- A website adds one line to its pages.
- That line loads a small JavaScript file from Optimizely's servers, called the snippet.
- Each customer gets their own snippet, containing Optimizely's code plus that customer's tests.
- When a marketer edits a test in the dashboard, the file changes. The website's developers don't have to touch anything.
That convenience is the selling point, but it's also the risk. To swap a button before you see it, the snippet must run very early. Optimizely's documentation tells customers to load it as high as possible in the page head, before the rest of the page. Otherwise visitors see the old button flash and then change, which is called "flicker."
So this one file sits at the top of the page, runs before almost everything else, can change anything you see, and can be edited from a dashboard without the site's own code changing.
Optimizely also has a setting called project JavaScript, which lets customers run extra code before the main snippet. It's essentially a text box, and whatever you type goes into that file and (according to Report URI) runs on every page.
That text box is where this attack lived.
How the day unfolded (UK time)
- 2:21 AM: An internet scanning service captured the BBC's Optimizely file. It was clean.
- 4:21 AM: The file had changed. A few new lines in the project JavaScript box told every visitor's browser to fetch one more script from a site the attacker controlled. The rest was untouched, and it still came from Optimizely's servers with Optimizely's security certificate. In Report URI's words, the delivery network served exactly what it was given. The bad code was planted at the source.
- 5:23 AM: The attacker registered a new web address, used for the rest of the day.
- Late morning: Zoom's file was changed, followed by Salesforce, CNN, Fox News, Air France, KLM, GitLab, Smartsheet, and Stamps.com.
- Through the day: Report URI counted 26 builds of the main script. The attacker kept adding features and new crypto wallets.
- 1:43 PM: A card skimmer was added.
- Between 5 and 6 PM: A third component appeared, aimed at iPhones and iPads.
- Evening: Report URI ran its sweep (1,611 files, 188 infected), and Helme posted a public warning about a breach of a third-party JavaScript host affecting hundreds of high-value sites.
Cleanup was fast. Ledger's, Smartsheet's, and the BBC's files were cleaned within minutes of each other around 9 PM. By 9:32 PM only 41 of 188 were still infected, and by 9:52 PM just one remained. Report URI says that from 11:11 PM, every infected Optimizely file was clean and has stayed clean. Step Security estimates most sites served the code for about 16 hours.
There was one worrying detail. Smartsheet's file was cleaned, then infected again, this time with the script's address scrambled into a different text format. Whoever wrote the code still had access.
On some sites, the attacker didn't even rely on the outside script. The BBC's file and one of HSBC's carried a full 24-kilobyte copy of the malicious code packed inside the Optimizely file. If the main script was blocked or took more than about 2.5 seconds to load, the built-in copy ran instead. Hudson Rock says this meant blocking the attacker's main domain wasn't enough.
What the code actually did
According to Report URI, every visitor got the same script, and all four of its jobs were active at once.
Job one: swapping crypto addresses. Nobody types a wallet address by hand, so you copy and paste it. The script hunted for wallet addresses and quietly replaced them with the attacker's, in page text, form fields, links, your clipboard, and outgoing requests. The built-in checksum doesn't help, because the attacker's address is a perfectly valid address. It just belongs to someone else.
Job two: stealing card numbers. Added in the afternoon, this part watched forms and outgoing requests for card number, expiry date, and security code. It had a blind spot, though. Many shops use payment boxes from providers like Stripe or Adyen, which are essentially windows into the provider's own site. A script on the shop's page can't read inside them. Cards typed into a site's own form fields were not protected.
Job three: screenshots and cookies. When the script swapped an address near words like "wallet" or "payment," it also sent back the page address, some cookies, and a screenshot. Cookies that sites had locked against script access were safe, according to Report URI.
Job four: targeting iPhones. On an iPhone or iPad, the script checked the iOS version, loaded a hidden frame, and tried to break into the device using exploit kits called Coruna and DarkSword. Google's threat intelligence team wrote about these earlier this year. Coruna covers iOS 13 up to 17.2.1 with 23 exploits, and Google watched it pass from a surveillance vendor's customer to spies to scammers. DarkSword targets iOS 18.4 to 18.7, and a newer version was reportedly leaked online.
What stands out to me is that tools once reserved for targeted individuals ended up in a script on news sites, banks, and shops, aimed at anyone with an older iPhone.
Every bug these kits use has been patched, in iOS 18.7.3 and iOS 26.3. For newer iPhones, the script only collected device details. There is no public proof that any real iPhone was compromised in this attack. Researchers tested with tools that imitate iPhones, and Hudson Rock says what was shown was delivery, not demonstrated compromise. One visible clue: on iOS 18.6 or older, the page reloaded itself every 60 seconds.
How much did the attacker make?
Only the crypto part is measurable, because crypto payments are public. Trackers disagree:
- Report URI counted about $7,040 across the attacker's wallets.
- Step Security counted about $6,600.
- Cointelegraph's check of the chain found about $7,100.
- Alex Thorne of Galaxy Digital suggested only about $2,300 came from real victims, with much of the Bitcoin looking like the attacker moving their own funds around.
So from crypto, it's a few thousand dollars, maybe less. The attacker emptied the wallets early Tuesday, and one cash-out went to an address Report URI says was used in earlier operations in March 2026 and on November 3, 2026.
The real cost is invisible. Nobody can say how many cards were skimmed, how many cookies or screenshots were stolen, or how many devices were hit.
How did the attacker get in?
Optimizely's status page said an unauthorized party exploited a vulnerability in its experimentation product and added malicious JavaScript to some customer projects. In a statement to Bleeping Computer, the company said the weakness was in the experimentation account signup flow. It says it blocked that path, ended active sessions, revoked access tokens, reset access for affected accounts, and notified customers.
The open question is how a flaw in signing up for a new account allowed someone to edit other companies' projects. Hudson Rock, Step Security, and Report URI all list this as unexplained, and as of Tuesday there was no full write-up from Optimizely.
Optimizely also wasn't the only door. Starting on the 13th, Hudson Rock found similar code in scripts from three smaller services: Kissmetrics, a story widget called Storyful-style plug-in "StoryifyMe," and a Minute Media video script. Through these, the code reached a Philippine bank, the Frankfurter Allgemeine, the Sydney Morning Herald, Vita XL, and Fox Sports. A shared tracking ID led one researcher to believe one buyer of the kit ran every wave. Hudson Rock is more careful: shared code links them, but that doesn't prove one operator or one entry method.
As for who is behind it, nobody knows. There are hints, such as a Chinese-language control panel and servers in Hong Kong and Singapore, but researchers stress these are easy to fake.
Why didn't security protections stop it?
Many big sites use a Content Security Policy (CSP), a list telling the browser which sources may run scripts and receive data. Report URI checked the affected sites:
- 68 had no CSP at all.
- 50 had one, but the attack still ran and sent data out.
- 17 let the script run but blocked data from leaving, so address swapping still worked.
- Only 24 blocked it completely.
The core problem is that a CSP checks where a script comes from, not what's inside it. This code came from Optimizely's own domain, which every site had approved. As Report URI put it, allowing the Optimizely domain was never going to help.
The older tool Subresource Integrity (SRI) does check contents: if a single byte changes, the browser refuses the file. But it only works for files that never change, and an Optimizely snippet changes whenever a marketer edits a test. Even Optimizely's own integrity guide fingerprints only the fixed part of the file.
So the file that can change anything on your page is also the hardest to lock down.
We've seen this before
- 2018, Ticketmaster UK: A chatbot from an outside vendor on its payment page was altered. The UK regulator fined Ticketmaster £1.25 million, with 9.4 million customers potentially affected.
- 2024, Polyfill.io: A free script service changed owners and was found injecting malicious code into more than 100,000 sites, according to Sansec.
Each time, a site trusted someone else's code, that code changed, and visitors paid. Since March last year, the payment card industry's standard requires shops to inventory scripts on payment pages, justify each one, and get alerted when one changes. But those rules cover only payment pages, not an A/B testing script on a homepage or news article.
What you should do
If you're a regular user:
- If you copied a crypto wallet address on an affected site on Monday, check your recent transfers.
- Before sending crypto, compare the full address, not just the first and last few characters. If you use a hardware wallet, verify on the device's own screen.
- If you typed a card number into an affected site on Monday, watch your statements and call your bank if anything looks off.
- Update your iPhone or iPad. Anything on iOS 18.7.3 or 26.3 or newer has every fix these kits target.
If you run a website:
- Inspect what is inside your third-party scripts, not just where they come from.
- Rotate access credentials for tools like Optimizely, and use strong authentication.
- Use a tight CSP, avoid
unsafe-evalwhere possible, and turn on CSP reporting so visitors' browsers tell you when something odd loads.
My takeaway
We tend to think of a website as one thing: one company, one address, one padlock. But the BBC's homepage on Monday was dozens of companies' code stitched together in your browser. According to the HTTP Archive, the biggest sites load a median of 66 outside services, and any one of them can change the page.
This time, one text box inside one vendor reached 178 organizations in a single day. The researchers who caught it were watching what the code did, not only where it came from. I think that is the lesson: trust in a source is not the same as safety of its content.
So yes, the headline is true. One attack reached 178 organizations through a single file they all trusted.