BBC.com Poisoned Script: How to Check Your iPhone and Cards
Who is affected: anyone who opened bbc.com, Zoom's website or a site belonging to one of at least 178 organisations in a browser during the hours its script was tampered with on 16 November 2026. The exposure window varied from site to site. For bbc.com it lasted from roughly 04:21 to 21:07 UTC.
The one-minute check: on an iPhone or iPad, open Settings > General > About and look at the iOS Version line. iOS 27, iOS 26.3 or later, and iOS 18.7.3 or later carry Apple's fixes for the known flaws the exploit kits rely on.
The fix: install the newest iOS via Settings > General > Software Update, then restart the device. If you entered a card number or sent cryptocurrency on an affected site that day, review your statement and the destination address.
Many Le VPN customers connect in order to reach the BBC. The site named by the researchers is bbc.com. BBC iPlayer, the service people actually watch through, appears in no source.
What happened on bbc.com on 16 November 2026
Optimizely is a testing tool that websites load when they want to compare different versions of a page. On 16 November 2026, someone inserted malicious JavaScript into the Optimizely projects of at least 178 organisations. From that point on, each site handed the code to its own visitors as part of the ordinary page.
Report URI fetched 1,611 snippets and found 188 infected ones belonging to 178 organisations, and says the real total is probably higher. The sites it names include the BBC, Zoom, CNN, Salesforce, GitLab, IKEA, HSBC, Equifax, Air France-KLM, Ledger and Peacock. It also names Fox News, NPR, Virgin Media, Vodafone Australia, Paychex and Starbucks Canada, plus HSBC sites in 11 markets. It counts 32 of the organisations in banking and finance.
Report URI's campaign timeline, all in UTC, with the BBC entries marked:
- 02:21: the BBC snippet is still clean.
- 04:21: the first injection appears in the BBC snippet, which starts loading an earlier injector (webshell.lol).
- 08:28: the stealing script from 1688360.cc is first seen on bbc.com.
- 11:20: Zoom's snippet is modified.
- 13:43: across the campaign, a card skimmer is added to the payload.
- 17:13 to 17:57: across the campaign, a second iOS host is added and a hidden frame for the Coruna exploit kit is injected into every visit that reports iOS 18.6 or earlier. Report URI does not give the time the iOS loader script first ran.
- 20:40 to 21:10: a mass cleanup. The BBC snippet is cleaned at 21:07.
- 21:52: one of the 188 snippets (TalkTalk) is still infected.
A cleanup did not always stick. Report URI notes that Smartsheet's snippet was infected again after a first cleanup.
Only bbc.com is named. No source names bbc.co.uk, BBC iPlayer, BBC Sounds or the BBC apps. As of 17 November 2026, according to Report URI and BleepingComputer, the BBC and Zoom had not commented.
Optimizely's status page says an unauthorized party exploited a vulnerability in its Experimentation product to add malicious JavaScript to some customer projects. It says the vulnerability is fixed, the code is removed and the affected accounts are secured. BleepingComputer reported that Optimizely placed the flaw in its account signup flow. The status page does not give that detail.
The code was served to every visitor of an affected page, in any browser. The researchers describe no targeting by country or account.
What the script did
It came in two parts. The first ran on every device. The second went after Apple devices only.
On every device
According to Report URI and StepSecurity, the payload did three things:
- Crypto address swapping. Any valid wallet address (Bitcoin, Ethereum, TRON, Solana and others) was replaced with the attacker's, in the page, in forms, links and images, in the clipboard, in outgoing requests and, on iOS, in the share sheet. Report URI says it can affect copying from any page while the tab is open.
- Card skimming. It read the card number, expiry date and CVV from payment fields that belong to the site itself. Payment forms hosted by a payment provider in their own frame were out of its reach.
- Cookie and screenshot theft. When it swapped an address next to wallet or payment wording, it sent the page address, the cookies that scripts can read and a screenshot to the attacker's server.
Report URI counted about USD 7,040 received in the attacker's wallets by 19:25 UTC on 16 November 2026.
On iPhones and iPads
Between 17:13 and 17:57 UTC, the script injected a hidden exploit frame for browsers that identified as an iPhone, iPad or iPod, or an iPad reporting as a Mac, and that reported iOS 18.6 or earlier. The time iPhones first received the iOS loader script has not been published.
Safari 26 on iOS reports "18_6" in its identification string whatever the real version, as WebKit documents. No researcher has said this, but on our reading of the WebKit note, up-to-date iPhones probably received the frame too. The known flaws the frame leads to are fixed on those versions.
Hudson Rock and the researcher Matt Kirkland traced the frame to a router that sorts devices by version. iOS 13 to 17 went towards an exploit kit called Coruna, and iOS 18 towards one called DarkSword. Google's threat intelligence group described both kits in March 2026: Coruna covers iOS 13.0 to 17.2.1 and DarkSword covers iOS 18.4 to 18.7. Apple has fixed every flaw Google lists for them, the last one in iOS 26.3 in February 2026. These were known, fixed bugs, not new ones.
For iOS 26, Report URI says the chain "currently only fingerprints the device". Kirkland tested on iOS 26.3 and found that, according to the chain's own log, the code that branch loaded was not weaponized.
Hudson Rock says its observations were made in a test browser, and that they show delivery, not proof that real iPhones were exploited. No researcher, Apple or any affected site has confirmed a compromised device. What the iOS stage would have installed in this campaign is unknown. In earlier campaigns, Google saw DarkSword deliver tools that copy messages, photos, saved passwords and crypto wallets, and Coruna deliver a loader aimed at wallet apps and seed phrases.
iOS versions: which are affected and which are safe
Open Settings > General > About and read the iOS Version line, then find your row in the table.
| Your iPhone or iPad runs | Status | What to do |
|---|---|---|
| iOS 27 (any version) | Known flaws fixed | Stay current. The latest is iOS 27.2.1. |
| iOS 26.3 or later | Known flaws fixed | Stay current. The latest is iOS 26.7.3. |
| iOS 26.0 to 26.2 | Vulnerable range | Update now. |
| iOS 18.7.3 or later | Known flaws fixed | Stay current. The latest is iOS 18.7.12. |
| iOS 18 below 18.7.3 | Vulnerable range | Update now. |
| iOS 17.3 or later | Above Coruna's stated range (iOS 13.0 to 17.2.1) | Update to the newest version offered. |
| iOS 17 below 17.3 | Vulnerable range | Update to the newest version offered. |
| iOS 16.7.15 or later, on a device that cannot leave iOS 16 | Carries Apple's Coruna fix | Install every update offered. |
| iOS 16 below 16.7.15 | Vulnerable range | Update now. |
| iOS 15.8.7 or later, on a device that cannot leave iOS 15 | Carries Apple's Coruna fix | Install every update offered. |
| iOS 15 below 15.8.7 | Vulnerable range | Update now. |
| iOS 14 or iOS 13 | Inside Coruna's stated range | Update if the device allows it. |
| iOS 12 or earlier | Below the stated range of both kits | Update if the device allows it. |
The 18.7.3 and 26.3 thresholds come from comments in the exploit kit's own code, as read by Report URI. Apple's notes for iOS 18.7.3 and iOS 26.3 list fixes for flaws that DarkSword uses. Apple's own advice is shorter: install the latest version. The current list is on Apple's security releases page.
For iOS 18 devices, Apple widened the availability of iOS 18.7.7 on 1 April 2026 so that users with Automatic Updates get protection against the DarkSword web attacks.
Checking your iPhone or iPad
An ordinary user cannot open an iPhone and look for this kind of implant. iOS has no built-in scanner for it. iVerify, a mobile security firm, says the DarkSword implants "do not contain a persistence mechanism". The attack disengages after it has taken data. The checks below tell you whether you were at risk. They do not give a clean bill of health.
- Check the version. Settings > General > About > iOS Version. Compare it with the table above. If you were on a "known flaws fixed" version on 16 November 2026, the Coruna and DarkSword flaws that Google and Apple have documented were already patched on your device. What the iOS stage delivered in this campaign has not been analysed, so that is the limit of what can be said.
- Update. Settings > General > Software Update. Then open Automatic Updates on the same screen and switch everything on, including Security Responses & System Files.
- Restart after updating. iVerify says DarkSword implants do not contain a persistence mechanism, so a restart is harmless and sensible. No source confirms that it removes whatever this campaign may have dropped.
- Think back to 16 November. Report URI noted one visible sign: on iOS 18.6 or earlier, affected pages reloaded every 60 seconds. If bbc.com kept refreshing on an old iPhone on 16 November, treat the device as exposed.
- Look for unknown profiles. Settings > General > VPN & Device Management. Remove any profile you did not install yourself. A work phone will show its employer's management profile, which is normal. No source says this campaign installs profiles, so this is general hygiene.
- Run Safety Check. Settings > Privacy & Security > Safety Check. It reviews who has access to your data and devices and can sign out other sessions. It does not scan for spyware.
- Check your Apple Account. Settings > [your name] > Sign-In & Security, and the device list under your name. Remove any device you do not recognise.
- Look at crash logs, with caution. Settings > Privacy & Security > Analytics & Improvements > Analytics Data. Repeated crashes of WebKit or wifid may be a weak signal, according to iVerify. One crash means little.
- Know what Apple threat notifications are for. Apple warns users on the Lock Screen, by email from threat-notifications@email.apple.com and at account.apple.com when it believes they were individually targeted by mercenary spyware. A mass campaign like this one may never trigger one. No notification is not an all-clear. Apple never asks you to click a link, install anything or give your password in such a message.
For people at higher risk
Journalists, activists and anyone holding large crypto balances on the phone should turn on Lockdown Mode, and can have a backup checked with Amnesty's Mobile Verification Toolkit.
Lockdown Mode cuts off both kits as the researchers describe them. Google found that Coruna stops when Lockdown Mode is on, and iVerify says DarkSword would not work under it without extra bypasses. The path is Settings > Privacy & Security > Lockdown Mode. It needs a restart.
The Mobile Verification Toolkit checks an encrypted iPhone backup against published indicators, including sets for DarkSword and Coruna. It is a command-line tool for technical users. Its indicators cover earlier campaigns and do not include this campaign's domains, so a clean result does not cover this campaign.
What to do if you were exposed, or cannot rule it out
Report URI's post lists the sites it names. If your iPhone was in the vulnerable range and you visited one of them on 16 November 2026, treat it as exposed:
- Update iOS, then restart.
- Turn on Lockdown Mode.
- From a different device you trust, change your Apple Account password, then your other important passwords.
- Remove unknown devices from your Apple Account.
- If you keep crypto keys or seed phrases on the phone, create a fresh wallet on a clean device and move the funds.
- If you work in journalism, activism or human rights, Access Now's Digital Security Helpline is free and open around the clock.
Checks for any visitor, on any device
No exploit for Android or desktop browsers was reported. The card skimmer, the address swapper and the cookie theft ran in every browser, so this part applies to Android phones, Windows and Mac computers, and iPhones alike, for anyone who used an affected site on 16 November 2026.
- Crypto transfers. If you copied, pasted or typed a wallet address on an affected site, or copied one anywhere in the browser while an affected page was open, open the transaction and compare the destination with the address you meant to use.
- Cards. If you typed a card number into a form on an affected site, watch the statement.
- Sessions. Sign out of the site and sign back in. That replaces the session cookies the script could read.
What a VPN did and did not do
A VPN did not stop this. A VPN encrypts the connection between your device and the internet. This script arrived from cdn.optimizely.com or, for some sites, from the site's own domain, with a valid certificate, as part of the real page. The tunnel carried it like any other part of bbc.com.
Reading bbc.com through a VPN neither added to your exposure nor reduced it. The code went to every visitor, and no source describes targeting by country.
Le VPN's Threat Protection is DNS filtering through Quad9, which blocks known malicious domains before they load. On the iPhone it is always on with IKEv2 and IPSec. On WireGuard the app sets Quad9 (9.9.9.9) with Cloudflare (1.1.1.1) behind it, unless the Ad Blocker is on.
Le VPN checked the campaign's published domains against Quad9 on 17 November 2026 at 06:33 UTC. That was after the cleanup. What Quad9 blocked while the code was live on 16 November is unknown. The result:
- Quad9 blocked six of the campaign domains.
- Quad9 did not block four others at that time, including the main router for mobile devices. Rechecked at 06:49 UTC, that router was still not blocked.
- The payload domain, 1688360.cc, had been suspended by its registrar and no longer resolved anywhere.
- Cloudflare's 1.1.1.1 does no filtering by design.
Hudson Rock also found that the bbc.com snippet embedded a full copy of the stealing code, which ran if the remote copy failed to load within 2.5 seconds. Report URI confirms this inline copy, and says that at HSBC the site's own security policy blocked the remote script and the data upload but the inline copy still ran. DNS filtering could not stop that code from running in the page. At most it could block the upload of what the code collected.
What did block the script was a setting on the website's side. Report URI found that the sites with a strict Content Security Policy refused to run it. That was 24 of 159 organisations, or 15%. Another 17 blocked the data upload but not the address swapping. Allowing cdn.optimizely.com in the policy did not help, because the code came from there.
About the author
Le VPN News Desk
The Le VPN Research Team is the news desk of the Le VPN blog. It tracks internet shutdowns, censorship, new privacy laws, and cybersecurity incidents as they unfold, and turns them into clear, sourced reporting. Every article is built from primary sources, fact-checked against them, and reviewed under Le VPN's editorial rules before it is published.
Articles by Le VPN Research Team →