Get a free audit
Apple credit Apple credits the researcher who runs our AI penetration testing.

Field Notes / Supply Chain

Zoom, the BBC and 32 finance brands served malware for a day. The code came in through their A/B testing tool.

On 16 November an attacker wrote malicious code into Optimizely Web Experimentation, and 178 organisations sent it to their visitors from a trusted Optimizely address. It swapped crypto addresses, skimmed cards and tried iPhone exploits. Here is what happened, which defences held, and what to change on your own site.

Author
Red Team Partners
Read
7 MIN READ
Filed
18 Nov 2026
A woman working on a laptop in a dark office, the screen the only light.

At 02:21 UTC on Monday 16 November, the Optimizely snippet on the BBC’s website was clean. At 04:21 it carried a loader for someone else’s code Report URI . By the evening, Report URI had found the same injection in 188 of the 1,611 Optimizely snippets it could fetch, belonging to 178 organisations, Zoom, CNN, Salesforce, IKEA, HSBC, Equifax, Air France-KLM and Ledger among them. Not one of those companies needed to be hacked. They were running a testing tool they trusted.

That is the whole story in two sentences, and it applies to almost every commercial website on the internet.

Nineteen hours on 16 November

Optimizely Web Experimentation is an A/B testing product. A site adds one script tag, and marketing can change headlines, buttons and layouts without a developer. BuiltWith counts roughly 190,000 live sites using Optimizely, though that is overall usage, not the number affected Malwarebytes . The attacker wrote code into each customer’s “Project JavaScript”, the field where customers add their own custom code. The rest of the snippet was untouched, and it still came from cdn.optimizely.com with Optimizely’s certificate Report URI .

Time (UTC, 16 Nov)What Report URI saw
04:21First Optimizely injection, in the BBC’s snippet
08:28The main payload, track.js, loads on bbc.com
11:20Zoom’s snippet modified
13:43A card skimmer is added to the payload
19:25About $7,040 in crypto counted in the attacker’s wallets
20:40 to 21:10Mass clean-up of infected snippets
21:52One snippet still infected, TalkTalk’s
23:11All monitored Optimizely files clean

The sector split is what made the story spread. Banking and finance was the largest group with 32 sites, ahead of technology at 23 and retail at 20. Eleven of the 32 are HSBC sites, in the US, Hong Kong, Mexico, Singapore, Australia, the UAE and five other markets Report URI . Optimizely’s own homepage and login page carried the code too.

Report URI says the real number is likely higher, because it could only check the snippets it found in time.

What the script did to visitors

Scott Helme of Report URI called it “a multi-purpose kit” The Record . The payload, rebuilt 26 times during the day, did four things Report URI :

  1. Swapped crypto addresses. Copy a Bitcoin, Ethereum, TRON, Solana, TON, Litecoin, Bitcoin Cash or Dogecoin address on an infected page and you got the attacker’s instead.
  2. Skimmed cards. From 13:43 it read card number, expiry and CVV from the site’s own form fields. Cards typed into a hosted payment iframe, such as Stripe or Adyen, were out of its reach.
  3. Took cookies and a screenshot. Any cookie without the HttpOnly flag went to the attacker, with the page URL and an image of the screen.
  4. Went after iPhones. On iOS it opened a hidden frame that loaded the Coruna kit for iOS 13 to 17 and DarkSword for iOS 18. Versions 18.7.3 and later, and 26.3 and later, are patched.

Two cautions on the scary parts. No source reports a confirmed iPhone takeover Security Affairs , and nobody knows how many cards were skimmed. Helme says there is no way to see how many cards, cookies or devices were hit before the attack stopped The Record . The roughly $7,000 in crypto is the only loss anyone has counted. (StepSecurity puts it nearer $6,600 Security Affairs .)

How it got in, as far as anyone knows

Optimizely’s notice, posted at 01:02 UTC on 17 November, says “an unauthorized party exploited a vulnerability in Experimentation and used it to add malicious JavaScript to some customer projects”, and that the company has “fixed the vulnerability and removed the injected code” Security Affairs . On X it added that “the vulnerable path is blocked” and that customers had been contacted directly The Record . It has not said what the flaw was.

Report URI rules out cache poisoning: Optimizely’s CDN “served exactly what it was given” Report URI . Somebody had write access to many customers’ projects at once. Smartsheet’s snippet was cleaned and then re-infected, so that access outlasted the first clean-up. The same injector had already appeared on Turkish sites from 17 September and in scripts from Kissmetrics, StorifyMe and Minute Media in the days before.

Which sites were protected, and why

Report URI checked the Content Security Policy on the affected organisations’ own sites. A CSP is a header that tells the browser which sources a page may load scripts from and send data to Report URI .

What the site hadSitesResult
No CSP68Everything ran
A CSP that allowed the payload and data theft50Everything ran
A CSP whose connect-src blocked the data theft17Payload ran, address swapping still worked
A CSP that blocked the payload24Protected

Allowing cdn.optimizely.com “was never going to help”, because the bad code came from there. The sites that held had a script-src listing only the hosts the page needed, so the loader could not fetch track.js from the attacker’s domain.

HSBC shows how close the margin was. Its CSP blocked track.js and the data theft, but an inline copy of the payload still ran, because the policy allowed 'unsafe-eval'. Optimizely’s own documentation tells customers to allow it Report URI . Report URI also warns about 'strict-dynamic', which trusts any script a trusted script inserts.

Subresource Integrity, the other textbook answer, would not have saved anyone here. It pins a script to a fixed hash, and an experimentation snippet is built to change every time marketing edits a test.

Four changes for your own site

If you run Optimizely, Report URI’s advice is direct: search your Project JavaScript and custom experiment code for 1688360.cc, atob( and anything you did not write, review every user, API token and integration with write access, treat cards entered on 16 November as possibly exposed, and rotate any secret held in a non-HttpOnly cookie Report URI . For everyone else, the work is the same with a different vendor name on it.

  1. Make a list of every third-party script on your payment and login pages, who owns each one, and who at the vendor can change it. Expect to find tags nobody remembers adding.
  2. Take testing and analytics tags off the checkout, or move card entry into a hosted payment iframe. The skimmer could not read Stripe or Adyen frames.
  3. Ship a strict CSP with reporting switched on: named hosts in script-src, a tight connect-src, no 'unsafe-eval' where you can avoid it. Reports would have shown 1688360.cc within minutes.
  4. Treat vendor admin access like your own. Single sign-on, MFA, named accounts, short-lived API tokens and a log of who changed what in the tool.

If you take cards, the first three are already on your PCI DSS list. Requirements 6.4.3 and 11.6.1 expect payment page scripts to be “properly authorized, checked for integrity, and monitored for tampering”, and changes to scripts and security headers to be detected PCI SSC . An A/B testing tag on the checkout counts.

The attacker reached 178 brands through one vendor and one field that customers trusted, and the same route into your site is a pasted tag. Our penetration testing covers the scripts your pages load and the accounts that can change them, the same way we would test a poisoned update channel or a hijacked CLI release. Start with your checkout page, and ask your marketing team which tags it loads.

Questions readers ask

What happened in the Optimizely attack in November 2026?

On 16 November 2026 an attacker added malicious JavaScript to customer projects in Optimizely Web Experimentation, a tool that lets websites run A/B tests. Because sites load Optimizely’s snippet on every page, the code reached visitors of 178 organisations, according to Report URI, which found it in 188 of 1,611 snippets it checked. Optimizely said an unauthorised party exploited a vulnerability in Experimentation, and that it fixed the vulnerability and removed the injected code.

Which companies were affected by the Optimizely malware?

Report URI names Zoom, the BBC, CNN, Salesforce, GitLab, IKEA, HSBC, Equifax, Air France-KLM, Ledger and Peacock among 178 organisations. Banking and finance was the largest sector with 32 sites, eleven of them HSBC sites in markets including the US, Hong Kong, Mexico, Singapore, Australia and the UAE. Report URI says the real number is likely higher because it could only check the snippets it found in time.

Were the affected companies hacked?

Their own servers were not shown to be breached. The malicious code sat inside each company’s Optimizely project and was served from cdn.optimizely.com with Optimizely’s certificate, so browsers treated it as the trusted script the site had asked for. Malwarebytes notes that the BBC example does not show the BBC’s own servers were breached. The company still served the code to its visitors and owns the consequences.

What did the malicious Optimizely script do?

Report URI describes four functions: it swapped cryptocurrency wallet addresses for the attacker’s, skimmed card numbers, expiry dates and CVVs typed into a site’s own form fields, sent non-HttpOnly cookies and a page screenshot to the attacker, and loaded the Coruna and DarkSword exploit kits for older iPhones and iPads. About $7,040 in crypto reached the attacker’s wallets by the evening of 16 November. No source reports a confirmed iPhone takeover.

Would a Content Security Policy have stopped the Optimizely attack?

Only a tight one. Report URI found 24 affected sites whose CSP blocked the payload outright, while 68 had no CSP and 50 had a policy that let both the payload and the data theft through. Allowing cdn.optimizely.com did not help, because the bad code came from there. What worked was a script-src listing only the hosts a page needs, a strict connect-src, and no unsafe-eval.

Does PCI DSS cover third-party scripts like Optimizely?

Yes, on payment pages. PCI DSS v4 requirements 6.4.3 and 11.6.1 expect payment page scripts to be authorised, checked for integrity and monitored for tampering, and changes to scripts and security-relevant HTTP headers to be detected. A testing tool loaded on the checkout page is one of those scripts. Cards entered into a hosted payment iframe such as Stripe or Adyen were out of the skimmer’s reach.

References

Sources

  1. Helme, S. “World’s Largest Brands Infected With Malware Via Optimizely.” Report URI, 16 November 2026, updated 17 November 2026. blog.report-uri.com
  2. Paganini, P. “OPTIMJACK: BBC, CNN and Zoom Among Targets of Optimizely Script Hijacking.” Security Affairs, 17 November 2026. securityaffairs.com
  3. Greig, J. “Website management platform Optimizely confirms hackers put malicious code into customer sites.” The Record, 17 November 2026. therecord.media
  4. Arntz, P. “CNN, BBC, Zoom, and other sites served malicious code in supply-chain attack.” Malwarebytes, 17 November 2026. malwarebytes.com
  5. PCI Security Standards Council. “New Information Supplement: Payment Page Security and Preventing E-Skimming.” 10 March 2025. blog.pcisecuritystandards.org