Attackers planted malicious JavaScript in Optimizely Web snippets belonging to 178 organizations on 16 November 2026, among them Zoom, GitLab, Salesforce, the BBC, HSBC and Ledger, according to Report URI. The code skimmed cards, swapped crypto wallet addresses and targeted iPhones, all while loading from Optimizely's own CDN under a valid certificate. If your pages run Optimizely, your Optimizely malware check starts with the project's edit history for that day, and no one has explained how the attacker got in.
Key takeaways
- On 16 November 2026, Optimizely snippets of 178 organizations served malware through the customer-controlled Project JavaScript field, per Report URI.
- The CDN and certificate were legitimate, and self-hosted copies of the snippet carried the same code.
- One payload swapped crypto addresses, skimmed cards from site-owned fields, stole non-HttpOnly cookies and screenshots, and pushed exploit chains to iPhones.
- Most cleanup happened in a 30-minute burst that evening, so the projectJS revision log is your evidence; today's download is not.
- Optimizely has not explained how the attacker got cross-customer write access, and one project was reinfected after cleaning.
What happened with Optimizely on 16 November 2026?
Visitors to 178 organizations received attacker code on 16 November 2026, delivered inside the Optimizely Web snippet those sites load to run A/B tests. The named victims include Zoom, GitLab, Salesforce, the BBC, CNN, IKEA, HSBC, Equifax, Air France-KLM, Ledger and Peacock, and Report URI sorts the full list into 20 sectors.
Report URI pulled 1,611 distinct Optimizely snippets, most of them found through its own crawl data, and checked each one for the injected code. 188 were infected, and those files belonged to 178 organizations. The company expects the real number is higher, since it could only check snippets it found in time.
Finance took the biggest hit: 32 banking and finance organizations, then 23 technology and SaaS firms, 20 retail and fashion brands, 12 streaming services, 10 news outlets and 9 airlines or travel companies. Healthcare and industrial firms account for 8 each, telecoms for 7, and crypto, insurance and property for 6 each.
Optimizely itself was among the victims. The snippets for www.optimizely.com and login.optimizely.com carried the same injected code, so whatever access the attacker had reached the vendor’s own projects as well as its customers’.
How the day unfolded
Report URI’s timeline puts the last clean capture of the BBC’s snippet at 02:21 UTC and the first injection at 04:21. The attacker registered the main payload domain at 05:23, built the first version of the script at 08:10 and had it loading on bbc.com by 08:28. Zoom’s snippet carries a Last-Modified time of 11:20.
The afternoon brought new features. The attacker swept stolen crypto to exchanges at 13:03 and added a card skimmer at 13:43. Between 17:13 and 17:57 a second iOS exploit host came online, and every visit from iOS 18.6 or earlier got a Coruna iframe. Report URI’s sweep from 19:24 to 19:59 found 186 of 1,611 snippets infected.
So the window to audit runs from about 04:20 UTC on the 16th until your snippet was cleaned that evening. An earlier anomaly still deserves a look: Report URI ties the same actor to injections through other vendors’ scripts starting 13 November.
Why a trusted Optimizely script can still be malicious
The attacker never touched Optimizely’s core code. They wrote into Project JavaScript, stored as projectJS inside the snippet: a field in every Optimizely project where the customer adds custom code. Whatever sits in that field gets bundled into the snippet and runs on every page that loads it.
Report URI says the rest of each snippet was untouched. The file still came from cdn.optimizely.com under Optimizely’s certificate, from a host every affected site already trusted. The report rules out cache poisoning: Optimizely’s CDN “served exactly what it was given.” The delivery channel worked as designed; the content it delivered was the problem.
Self-hosting didn’t help. Some companies serve the Optimizely file from their own domain, and Report URI found the same malicious content at paths like opt.scheels.com/js/<id>.js and www.vodafone.com.au/js/<id>.js. Those copies come from the same Optimizely project, so they inherit whatever is in projectJS. Moving the file changes the hostname, not the source of truth.
That is also why a clean WordPress or Joomla install means little here. Unless you self-host the snippet, nothing on your server changed. The code lived in a dashboard field, ran in visitors’ browsers and sent its loot straight to the attacker. The hijacked third-party RUM script on Polymarket worked the same way.
How did the attacker get in?
Nobody outside Optimizely knows yet. Report URI lists the options without choosing: “a compromised Optimizely system, a compromised credential with cross-customer access, or something else. That’s for Optimizely to explain.”
One detail shows the access stayed live during the response. Smartsheet’s snippet was cleaned, then infected again with a base64-encoded loader, which Report URI reads as proof that whoever wrote the code could still edit projects while customers were cleaning up. If you fixed your project on the 16th, check it again today.
Report URI’s write-up includes no statement from Optimizely. If your revision log shows edits you can’t explain, open a ticket from your Optimizely account with the revision IDs, timestamps and the user behind each change, and ask whether your project falls inside the incident scope.
Your own logins are the part you control. If an agency, a marketing team and two former contractors all share one Optimizely account, fix that this week, whether or not it played any role here.
What did the Optimizely JavaScript malware do to visitors?
The injected loader fetched one payload that did four jobs at once, for every visitor on every page carrying the snippet. What happened to a given visitor depended on their device and what they did on the page. The attacker rebuilt the script 26 times that day and, per Report URI, “added features, and added wallets but never removed any.”
Crypto address swapping
The script hunted for wallet addresses on Bitcoin, Ethereum and other EVM chains, TRON, Solana, TON, Litecoin, Bitcoin Cash and Dogecoin. It validated checksums so it only replaced real addresses, and it re-scanned the page whenever the content changed.
The swaps reached page text, form fields, links, images, the clipboard, outgoing fetch and XHR requests, window.open calls, page navigation and the iOS share sheet. Copy an address from an affected page, or from anywhere else while that tab stayed open, and you could paste the attacker’s without noticing.
Card skimming
From 13:43 UTC the payload carried a skimmer. It watched form submissions and every fetch or XHR POST for a card number, expiry date and CVV, Luhn-checked the number to drop junk, and sent complete sets out with navigator.sendBeacon. That API exists so analytics can report as a page closes, which also makes stolen data look like routine telemetry in a network log.
The browser’s same-origin rule set the limit. Script in the merchant page can read fields the merchant owns, but not fields inside a hosted Stripe or Adyen iframe. Two shops on the same processor could have very different exposure depending on where the customer typed the card.
Cookies and screenshots
When a swap happened next to wallet or payment wording, the script sent home the page URL, document.cookie, the original addresses it replaced and a screenshot of the page drawn with html2canvas. Only cookies without the HttpOnly flag were readable. Treat any JavaScript-readable session cookie on such a page as copied.
iPhone and iPad exploits
On iOS the loader read the system version and opened a hidden iframe with a matching exploit chain: Coruna for iOS 13 to 17, DarkSword for iOS 18.x, and a separate chain for iOS 26 that, at the time, only fingerprinted the device. Comments in the kit itself say iOS 18.7.3 and later, and 26.3 and later, are patched.
On iOS 18.6 or earlier the page reloaded every 60 seconds to retry the exploit, a symptom visitors may have blamed on a glitchy site. The Coruna loader matched, byte for byte, a sample Report URI analyzed in October from a different campaign.
Capability isn’t confirmed loss. The report shows what the code could do; it confirms no count of stolen cards, cookies or compromised phones.
How much did the attacker make?
By 19:25 UTC the attacker’s wallets had received about USD 7,040, all within roughly nine hours on the 16th. The bulk was 0.083 BTC: $5,380 across four transactions, still unspent. The rest was 17.02 LTC ($762), 662.9 USDT plus 184 TRX on TRON ($719), 0.055 ETH ($100) and 82 USDT on BSC ($82).
That tally counts what arrived, not what stayed. The attacker swept funds out between 13:03 and 13:05 UTC, with USDC on Ethereum going to a HitBTC deposit address. The TRON wallet has been active since March 2026, so this actor was working long before this campaign.
Don’t read $7,040 as the full damage. Report URI can’t see how many cards or cookies reached the attacker’s server, and any money made from that data would never pass through these wallets.
Optimizely malware check: did my site load the bad snippet?
Start by confirming whether any page loads Optimizely at all. Search the source of your homepage, login, account and checkout pages for cdn.optimizely.com, then look for an Optimizely project file served from your own domain under a /js/ path.
On WordPress, check the theme’s header template, any header-and-footer scripts plugin and Google Tag Manager. On Joomla, check the template’s index.php and any custom HTML module that writes into the page head. Marketing teams often add Optimizely through Tag Manager without telling whoever runs the site, so it may not appear in your theme at all.
A keyword scan of today’s file proves nothing. Report URI saw three loader styles:
- a plain script element pointing at an attacker-controlled .cc domain;
- the same URL base64-encoded and decoded with atob(), so a text search for the domain misses it;
- a full inline copy of the payload, run through eval if the external host was blocked.
Timing matters more than content. Most snippets were cleaned between about 20:40 and 21:10 UTC. At 20:56 UTC, 169 were still infected; by 21:52 only TalkTalk’s remained. Ledger cleaned at 20:41, Smartsheet at 20:56, the BBC at 21:07. A clean download now says nothing about what visitors got at noon.
Collect the history instead:
- Optimizely revision log and account activity for projectJS around 16 November. Any edit you can’t tie to a named colleague is a red flag.
- Saved copies of the snippet from monitoring or synthetic tests, diffed against a known-good version. Report URI kept every infected file with headers, hash and fetch time; your own copies serve the same purpose.
- CSP reports from the 16th showing blocked or unexpected script, connect and frame destinations.
- Tag manager versions for the same window.
- Proxy or browser telemetry showing unfamiliar script, iframe or beacon destinations.
Origin server logs rarely help, because they don’t record what JavaScript in a visitor’s browser sent elsewhere. Report URI’s summary fits: a website can be compromised through a third-party dependency “without its own code or infrastructure ever being breached.”
What should a site owner do now?
Treat any Optimizely project as possibly exposed until Optimizely explains the access path, and work in this order:
- Preserve first. Save current CDN and self-hosted responses with headers, timestamps and hashes, and export the revision history. Wiping projectJS before you do that destroys your best evidence.
- Clean at the project layer. Restore projectJS to a known-good version inside Optimizely. Cleaning WordPress files does nothing if the code lives in the project, and a self-hosted copy needs refreshing once the project is clean.
- Cut access. Rotate Optimizely passwords and API tokens, revoke unknown sessions, trim edit rights and require multi-factor login.
- Verify the bytes for days, not once. Fetch the snippet as an anonymous browser from both the CDN and your domain, and diff it against your clean copy. Smartsheet’s project picked the code back up after its first cleanup.
- Size exposure. List affected pages, whether card fields sat in hosted iframes or in your own inputs, and which cookies lacked HttpOnly. If session cookies were readable, invalidate sessions.
- Notify. Shops with their own card fields should involve their payment processor. Crypto services should tell users to verify any address copied or pasted on the 16th.
Then write down who can edit every script vendor your pages load: marketing staff, agencies, former contractors. Keep a known-good copy of each script, and pair your own diffs with website antivirus monitoring so a changed file on your side gets flagged the same day.
Should you remove Optimizely? If no experiments are running, yes. Either way, keep testing tags off checkout, login and account pages. They have no reason to sit next to card fields.
If you visited an affected site on the 16th
A visitor has no reliable way to prove what their browser received; history only shows which sites you opened. Act on what the payload could do instead, and pass these steps to your own customers if your site was on the list:
- iPhone or iPad: update to iOS 18.7.3 or later, or 26.3 or later, the versions the kit’s own comments list as patched. A page that reloaded itself every minute on an older iPhone was the visible symptom.
- Crypto: check every transaction sent on the 16th against the address you meant to use, including addresses copied from other tabs.
- Cards: if you typed card details into a listed site’s own form after 13:43 UTC, watch the statement and ask the bank about a replacement.
- Accounts: log out of listed sites and sign back in to retire any session cookie the script could read.
Why CSP, SRI and HTTPS didn't stop this third-party script compromise
Most browser controls vouch for the delivery channel, and the channel here was legitimate. The padlock proved the file came from Optimizely, not who last edited the project.
Subresource Integrity pins a script to a fixed hash. An experimentation snippet changes with every test edit, so pinning means updating the hash each time, which few teams manage. A Content Security Policy that trusts cdn.optimizely.com can’t separate good project code from bad.
Tighter rules still remove stages of this attack. A strict script-src blocks the external loader, connect-src limits where beacons can go, frame-src restricts hidden iframes, and refusing ‘unsafe-eval’ kills the inline fallback. Report URI, which sells CSP monitoring, argues that a policy in enforcement mode could have blocked key parts of the attack.
On WordPress you can send a CSP header from the server config or a security plugin. Start in report-only mode for a week, read what it flags, then enforce. HttpOnly kept session tokens out of document.cookie, so set it on every session cookie you control.
For shops, the lesson covers every script on checkout: keep a list of what is allowed there and get alerted when any of it changes. An A/B testing tag has no business on that list, as the checkout skimmer hidden in a shared tracker at Convertr showed.
What we still don't know
The entry point is the biggest gap. Until Optimizely explains it, nobody can say whether the hole is closed or whether projects outside Report URI’s 1,611-snippet sample were touched. The report names no CVE identifiers for the iOS chains and gives no victim totals.
Report URI does tie this to earlier activity. It says the same actor’s injector first appeared on 17 September on Turkish sites through an SEO “backlink panel” script. A separate wave from 13 to 16 November pushed a loader through Kissmetrics, StorifyMe and Minute Media scripts onto sites including securitybank.com, faz.net, smh.com.au, vidaxl and foxsports.com.
That attribution is the vendor’s assessment. The method is consistent: find a script thousands of sites load, get write access, ride the trust. If you run Kissmetrics, StorifyMe or Minute Media, apply the same history check described above.
Hacker News commenters were split on the report’s writing quality, but few disputed the event. When Optimizely publishes its account, compare its project list and time window with your own history.
FAQ
Was Optimizely hacked on 16 November 2026?
Attacker code was written into the Project JavaScript field of 178 organizations' Optimizely projects, including Optimizely's own. Report URI says the CDN was not poisoned and does not know how the write access was obtained.
What is Optimizely Project JavaScript?
It is a custom-code field in each Optimizely Web project. Its contents are bundled into the snippet and run on every page that loads it, which is why the attacker targeted it.
How do I check if my website loaded the compromised Optimizely snippet?
Find every place your pages or tag manager load Optimizely, then pull the project's revision log and account activity for 16 November. Edits nobody on your team made are the signal you are looking for.
Were self-hosted Optimizely snippets affected?
Yes. Copies on customer domains carried the same malicious code because they are generated from the same Optimizely project.
I visited one of the affected sites. What should I do?
Update iPhones and iPads to iOS 18.7.3 or 26.3 or later, check crypto transactions sent that day against the intended address, watch card statements if you entered a card on the site's own form, and log out and back in.
Does a valid CDN certificate mean a script is safe?
No. The certificate proves where the file came from, not who edited its contents.
If you can’t tell which third-party scripts your pages load or who can edit them, a website protection package from SiteGuarding gives you monitoring and a cleanup path when one goes bad.
