
Optimizely Web served malware to 178 organisations: what is confirmed, what is not, and how to check your site
A trusted script turned into a card skimmer and an iPhone exploit loader. Facts from Report URI, Optimizely and Google, and a checklist for site owners.
On November 16, 2026, security researcher Scott Helme of Report URI published findings that malicious JavaScript was being served through Optimizely Web, a service many large sites use to run experiments on their pages. Report URI found 188 infected snippets belonging to 178 organisations, including Zoom, the BBC, CNN, Salesforce, GitLab, IKEA, HSBC, Equifax, Air France-KLM, Ledger and Peacock (Report URI).
This piece separates what Report URI measured, what Optimizely has said, and what is still unknown, then lists what a site owner and an iPhone user can do.
What happened
Optimizely customers load a JavaScript snippet from cdn.optimizely.com. The snippet can contain a Project JavaScript field where customers keep custom code that runs on every page. Report URI says the attacker wrote a loader into that field and left the rest of the snippet untouched, so the file was still served from Optimizely's trusted domain. Report URI also states this was not a CDN cache-poisoning attack: the CDN served what it was given.
Timeline (UTC, Nov 2026, as reported by Report URI unless noted)
- 04:21 earliest injection Report URI sawA BBC snippet loadingsource
webshell.lol. Report URI's own data. - 19:24–19:59 Report URI sweeps 1,611 snippets186 infected at first count, 188 in the final count.source
- 23:11 all infected Optimizely files cleanedFrom Report URI's Nov 17 update. One customer, Smartsheet, was reinfected after an earlier cleanup.source
- 01:02 Optimizely posts a status noticeOptimizely says an unauthorized party exploited a vulnerability in Experimentation to inject malicious JavaScript into some customer projects, and that it fixed the vulnerability and removed the code. See the status page.source
What the code did
Report URI describes four behaviours in the payload:
- Crypto address swapping. Replaces wallet addresses (Bitcoin, Ethereum and EVM chains, TRON, Solana, TON, Litecoin, Bitcoin Cash, Dogecoin) with the attacker's, in page content, the clipboard, outgoing requests and the iOS share sheet.
- Card skimming. Added at 13:43 UTC on Nov 16. Captures card number, expiry and CVV from forms and POST requests. Hosted payment iframes, such as Stripe or Adyen fields, were out of its reach. Report URI says it cannot see how many cards were captured.
- Cookie and screenshot theft. Sends the page URL, cookies that are not HttpOnly and a screenshot to the attacker's server when crypto or payment content is on the page.
- iPhone exploit kits. On iPhones and iPads it loads an exploit chain in a hidden frame: Coruna for iOS 13 to 17, DarkSword for iOS 18.x, and a third chain for iOS 26 that, per Report URI, currently only fingerprints the device.
Losses reported so far are small: about $7,040 across the attacker's wallets at 19:25 UTC on Nov 16, with a later update raising the Bitcoin and Litecoin totals. The attacker moved funds out at 13:03–13:05 UTC and emptied the wallets early on Nov 17. The loss from card data is unknown.
The iPhone part: what is confirmed
Coruna and DarkSword are two exploit kits that Google's Threat Intelligence Group (GTIG) documented earlier in 2026. They are not new, and they only work against iPhones that have not been updated.
- DarkSword supports iOS 18.4 through 18.7 and uses six vulnerabilities. GTIG says all of them were patched with the release of iOS 26.3, "although most were patched prior" (GTIG).
- The six: CVE-2025-31277 (fixed in iOS 18.6), CVE-2025-43510 (18.7.2 and 26.1), CVE-2025-43520 (18.7.2 and 26.1), CVE-2025-43529 and CVE-2025-14174 (both 18.7.3 and 26.2), and CVE-2026-20700 (26.3).
- Coruna covers iOS 13.0 through 17.2.1 with five full exploit chains and 23 exploits, per GTIG, and does not work against current iOS (GTIG).
- Report URI says that iOS 18.7.3 and 26.3 are patched. That matches GTIG for the 26 branch; for the 18 branch the fixes arrived across 18.6, 18.7.2 and 18.7.3, and we did not find a source that states which 18.x release closes the dyld flaw. The practical advice does not depend on the exact version: install the latest iOS update.
These are not new zero-days found by the Optimizely attacker. The kits were already in use by other actors: GTIG names suspected Russian espionage group UNC6353 as a user of both. What is new is the delivery route: a script on a major news or SaaS site, instead of a dedicated lure page.
Optimizely's explanation, and where it stops
- Optimizely's status page: an unauthorized party "exploited a vulnerability in Experimentation" and injected malicious JavaScript into some customer projects. Optimizely says it fixed the vulnerability and removed the injected code, secured affected user accounts and added protections against unauthorized code delivery. The page gives no count and no technical detail (status page).
- A spokesperson, quoted by iTWire from a Gizmodo interview, said the flaw was in the Experimentation account sign-up flow. We could not check this in a primary source.
- Report URI says it does not know how the attacker obtained write access to so many unrelated customers' projects, and that Optimizely should explain. It also reports no reply from Optimizely to its outreach at the time of publication.
Why a Content Security Policy mattered
A CSP is a header that tells the browser which scripts it may run and where a page may send data. Report URI crawled 256 hosts and sorted them by policy:
- No CSP at all: 68 organisations.
- Payload and data theft both allowed: 50.
- Payload ran, but
connect-srcblocked the data upload: 17 (address swapping still worked). - Payload blocked: 24, about 15%.
These four groups add up to 159, not 256, and Report URI does not explain the other 97, so read the figures as indicative. The article says a restrictive script-src separated the protected sites, and that 'strict-dynamic' and 'unsafe-eval' let the payload through. HSBC's CSP blocked the external loader and data theft, but a fallback inline copy still ran.
Optimizely's own support article lists 'unsafe-eval' as needed for some features, for example audiences with custom JavaScript conditions (Optimizely support). That is a trade-off site owners should weigh: the feature that needs it is also what lets injected code run.
What to do
If you run Optimizely Web
- Search every project's Project JavaScript and experiment code for
1688360.cc,webshell.lol,atob(and scripts you do not recognise. Report URI notes that fallback copies were embedded, so check what browsers actually receive, not only what the dashboard shows. - Audit and rotate Optimizely access. Review users, API tokens and SSO links; remove accounts you do not know.
- Treat card data entered on Nov 16 as possibly exposed if your checkout is on a page with the snippet; coordinate with your payment provider.
- Rotate session tokens and secrets that were visible to page scripts, and revoke non-HttpOnly cookies that carried sensitive data.
- Tighten CSP. Restrict
connect-srcso injected code cannot send data out, avoid'unsafe-eval'where you can, and use nonces with a restrictivescript-src. - Add a safety net for third-party scripts: Subresource Integrity cannot protect a script that changes by design, so monitor what each third-party script serves and alert on changes.
If you use an iPhone or iPad
- Update iOS. Install the latest version your device supports; DarkSword was only closed in full in iOS 26.3 per GTIG, and Coruna does not work on current iOS.
- If you cannot update, GTIG recommends Lockdown Mode for the Coruna kit.
- Check crypto transfers made on Nov 16 for a changed destination address, and watch card statements.
Was Optimizely's CDN hacked?
Is my iPhone at risk now?
How many sites were affected?
Were cards stolen?
Data as of 2026-11-17live record →
Sources: Report URI, World's largest brands infected with malware via Optimizely, Nov 16–17, 2026, Optimizely status page, Google GTIG, DarkSword, Google GTIG, Coruna, Optimizely support, CSP. Also referenced: iTWire for the spokesperson quote.