On Monday 16 November 2026, Report URI found malicious code in Optimizely snippets used on sites including the BBC, Zoom, CNN, Salesforce and regional HSBC sites. The code arrived through Optimizely Web Experimentation, the A/B testing tool their own teams had added to the page. This post sets out what happened, what Optimizely has and has not explained, and the checks to run if your site loads Optimizely or any other third-party script.
The answer in brief
- What happened. An attacker added malicious JavaScript to Optimizely Web Experimentation customer projects on 16 November 2026. Security firm Report URI found it in 188 of the 1,611 Optimizely snippets it checked, belonging to 178 organisations.
- Who was hit in the UK. Report URI’s list includes the BBC, TalkTalk, O2, Virgin Media, ASOS, Equifax UK, NOW TV, Admiral, interactive investor and Tombola. HSBC is listed through several regional sites.
- What the code did. It could swap cryptocurrency wallet addresses, skim card details typed into a site’s own forms, collect readable cookies and screenshots, and direct older iPhones and iPads to exploit kits.
- The cause. Optimizely says an attacker exploited a vulnerability in its Experimentation account signup flow, and that the flaw is fixed and the code removed. Its interim report says the flaw let the attacker impersonate customer administrators and edit their projects.
- What to do. Optimizely Web customers should review Project JavaScript and its change history, preserve logs and assess whether payment details or session cookies were exposed. Everyone else should audit which third-party scripts can run on their pages, starting with checkout.
What happened on 16 November 2026?
Malicious code appeared in Optimizely customers’ snippets on 16 November 2026. Scott Helme’s team at Report URI published a full analysis the same day. Its earliest observation is at 04:21 UTC, when the BBC’s snippet was loading an earlier loader from webshell[.]lol. A capture two hours before was clean. Optimizely’s interim report gives a later time. It says the first malicious code was written to a customer project at 07:51 UTC, a time it labels preliminary, and that the code was removed from all affected projects by 22:24 UTC.
Neither source explains the three and a half hours between those two times. Report URI’s post does not mention Optimizely’s time, and Optimizely’s report describes no change to a customer snippet before 07:51 UTC. Hudson Rock separately dates an archived capture of the BBC file, loading the same earlier loader, to 04:31 UTC. The main payload came later: Report URI puts its first build at 08:10 UTC and its first sighting on bbc.com at 08:28 UTC, and counted 26 rebuilds during the day.
Report URI says the true number of affected organisations is likely higher than 178, because it could only check the snippets it found in time. Optimizely’s own website and login page carried the code too.
If the table extends beyond the screen, scroll sideways to view all columns.
| Time (UTC) | Event |
|---|---|
| 13 to 16 November | EventRelated malicious code is observed in Kissmetrics and StorifyMe scripts and the minute-ly video loader. Report URI counted 178 organisations in its Optimizely sample. |
| 16 November, 02:21 | EventLast clean urlscan capture of the BBC’s Optimizely snippet, cited by Report URI |
| 16 November, 04:21 | EventReport URI’s earliest observation: the BBC’s snippet loads an earlier loader from webshell[.]lol |
| 16 November, 05:23 | EventThe attacker registers the domain that will host the main payload |
| 16 November, 07:51 | EventOptimizely’s preliminary time for the first malicious code written to a customer project |
| 16 November, 08:28 | EventReport URI first sees the main payload loading on bbc.com |
| 16 November, 13:43 | EventA card skimmer is added to the payload |
| 16 November, 17:13 to 17:57 | EventAn iOS exploit kit is added for every visit from iOS 18.6 or earlier |
| 16 November, 19:24 to 19:59 | EventReport URI’s sweep finds 186 of 1,611 snippets infected |
| 16 November, about 20:40 to 21:10 | EventMass clean-up of infected snippets |
| 16 November, 21:52 | EventOne infected snippet is left, TalkTalk’s |
| 16 November, 22:24 | EventOptimizely’s preliminary time for the code being removed from all affected projects |
| 16 November, 23:11 | EventReport URI finds all previously infected files in its sample clean |
| 17 November, 00:37 | EventThe payload domain is put on hold and stops resolving |
| 17 November, 01:02 | EventOptimizely posts its incident notice |
Sources: Report URI for its observations, Optimizely’s interim report for the 07:51 and 22:24 times, and the Optimizely status page for the notice.
How did the attackers get in?
The attacker exploited a flaw in Optimizely’s signup flow to impersonate customer administrators and change their projects. Optimizely Web usually delivers its JavaScript snippet through its content delivery network. The snippet can include custom code from a field called Project JavaScript, which runs on every page where that snippet is loaded. The attacker changed this field in affected projects, and Optimizely says the attacker’s code replaced any custom code already there.
The file was still served from cdn.optimizely.com with Optimizely’s valid certificate, so browsers and security tools treated it as trusted. Sites that serve the snippet from their own domain were hit as well, because the content comes from the same project.
Optimizely confirmed the incident in a status page notice at 01:02 UTC on 17 November. It says an unauthorised party “exploited a vulnerability in Experimentation” to add malicious JavaScript to some customer projects. It also says it has fixed the vulnerability, removed the injected code, secured affected user accounts and added protections against unauthorised code delivery. The notice itself lists no steps for customers. An update on 18 November says less than 5% of Optimizely’s customers were affected, that no other Optimizely product was involved, and that Optimizely has contacted every Web Experimentation customer to confirm whether they were affected.
In a statement to BleepingComputer, Optimizely placed the flaw “in the Experimentation account signup flow”. It said it had blocked the vulnerable path, ended active sessions, revoked tokens, reset access for affected accounts and notified all affected customers.
Optimizely’s interim incident report, dated 17 November and corrected on 18 November, adds what the notice leaves out. It says the flaw let the attacker act as administrator users on customer accounts and edit Project JavaScript, and that this was not a breach of customer credentials. It asks customers to investigate their exposure, review card and crypto transactions from the day and keep their logs. The report explains how the attacker impersonated customer administrators. The exact signup flaw remains unspecified, and Optimizely gives no count of affected customers or projects. A final root cause analysis is promised once the investigation ends, with no date given.
One detail suggests the attacker kept access during the clean-up: Smartsheet’s snippet was cleaned once, then re-infected with a disguised version of the loader. Some researchers call the campaign OPTIMJACK.
What did the malicious code do?
What a visitor encountered depended on when they visited, the page’s security policy and their device. The malicious code had four jobs:
- Swapping crypto addresses. It looked for valid wallet addresses in page text, forms, links and the clipboard, and replaced them with the attacker’s. Someone copying an address on an affected page could paste a different one.
- Skimming cards. From 13:43 UTC it watched forms for a card number, expiry date and security code. Cards typed into a site’s own fields could have been exposed. The skimmer could not read card details inside a payment provider’s cross-origin iframe, such as those used by Stripe or Adyen.
- Taking cookies and screenshots. After certain wallet address swaps, it tried to send the page address, a screenshot and any cookies JavaScript could read. Cookies marked
HttpOnlywere not readable. - Targeting iPhones and iPads. It loaded exploit kits that researchers identify as Coruna and DarkSword. Report URI documented an exploit page and 60-second reloads targeting visits from devices running iOS 18.6 or earlier.
Report URI recorded about US$7,040 in payments into wallets linked to the malware by 19:25 UTC on 16 November. It could not see how many cards were skimmed, how many cookies and screenshots were taken or how many devices were infected. Optimizely’s interim report says it has not confirmed whether any card details were captured, and that it can see when the code loaded but not whether it succeeded.
Two cautions apply to the scarier claims. Hudson Rock, a second firm that analysed the campaign, has not linked those wallet payments to visitors of the affected sites, and says it has not shown a real visitor’s iPhone being compromised. Being served an exploit page is not the same as being hacked.
Which UK organisations were affected?
Report URI’s published list includes these UK names and domains: the BBC (bbc.co.uk and bbc.com), TalkTalk, O2, Virgin Media, Virgin Media O2, ASOS, Equifax UK, NOW TV, Admiral, interactive investor, Tombola and Top Gear. HSBC appears with several regional sites, and told Reuters that unauthorised scripts were not active on its websites.
Banking and finance is the largest group on the full list with 32 organisations, followed by technology with 23 and retail with 20.
A place on the list means the organisation’s Optimizely snippet carried the code. It does not show that any visitor lost money or data. What a visitor was exposed to depended on the page: a news article with no payment form gave the skimmer nothing to read.
Why did a Content Security Policy not stop it?
Many sites had no Content Security Policy (CSP), or had one that still let the payload run. A CSP tells the browser which scripts a page may run and where it may send data. Report URI tested the affected organisations’ policies against what the payload needed:
If the table extends beyond the screen, scroll sideways to view all columns.
| Result | Organisations |
|---|---|
| No CSP at all | Organisations68 |
| CSP allowed both the payload and the data theft | Organisations50 |
| Payload ran, but the policy blocked data being sent out | Organisations17 |
| Payload blocked | Organisations24 |
Source: Report URI.
Of the 159 organisations in Report URI’s four rows, 24 (15%) blocked the payload. Allowing the Optimizely CDN by itself did not stop the injected code, because that code arrived inside the snippet. The 17 policies that blocked outgoing data could still let wallet addresses be changed on the page.
Two common settings made things worse. Optimizely’s CSP guide includes 'strict-dynamic', which can let a trusted snippet load additional scripts dynamically. 'unsafe-eval' let an inline copy of the payload run even where the external file was blocked. Report URI says that happened at HSBC, whose policy blocked both the external file and the data theft. The advice to allow 'unsafe-eval' comes from a separate Optimizely page, its Experimentation Assistant guide.
Subresource Integrity, the usual check for third-party files, can only verify a file that stays the same. A live Optimizely snippet changes whenever someone saves a change that goes into it, so the site would have to update the fingerprint after every approved edit.
What should Optimizely Web customers check now?
Check which pages loaded the affected snippet, review Project JavaScript and preserve the relevant logs. Optimizely says it removed the code and has contacted every Web Experimentation customer to confirm whether they were affected. Its status update of 18 November puts the share of its customers affected at less than 5%. Neither its notice nor its interim report gives a count of changed projects, and Report URI only tested the snippets it could find. The steps below draw on Optimizely’s interim report and the advice from Report URI and StepSecurity.
- Read every Project JavaScript field. Do the same for custom code inside each experiment. Search for
1688360, foratob(and for any script your team did not write. A search for the domain alone misses the disguised and inline versions of the loader. - Review the change history. Each project’s History log records changes to Project JavaScript and whether they came through the interface or the REST API. Look at 16 November first, then the days before it.
- Cut back who can write. List the users, API tokens and integrations with edit rights. Investigate unfamiliar access and remove it if unauthorised. Optimizely says customer credentials and API tokens do not need a general reset for this incident.
- Ask Optimizely for specifics in writing. Which of your projects were changed, at what times, and through which account. Keep the reply with your incident record.
- Work out what your pages exposed. If the snippet ran on a page with your own card fields, treat cards entered on 16 November as possibly exposed and speak to your payment provider. If session tokens sit in cookies without the
HttpOnlyflag, end those sessions and reissue them. - Search your logs. Look for requests to the attacker’s domains across the whole of 16 November (UTC), in proxy logs, DNS logs and CSP reports. Keep those logs: Optimizely asks customers to retain records covering at least 16 November.
- Decide whether to report. Under the UK GDPR, a data controller must report a personal data breach to the ICO without undue delay and, where feasible, within 72 hours of becoming aware of it, unless the breach is unlikely to put people’s rights and freedoms at risk. Your data protection lead should make that call and record the reasoning either way.
The domains to search for, written with brackets so they cannot be clicked. The first four come from Report URI. The last three are named in sections 7 and 9 of Optimizely’s interim report, which calls the first of them the skimmer’s exfiltration endpoint and the last a host for the skimmer:
1688360[.]cc
webshell[.]lol
5vobauua3ojpklx[.]icu
eil786npto9u5bq[.]icu
pay-relay.wordpressadmin001.workers[.]dev
pay-cloud[.]online
pay.pay-cloud[.]online
What should every website owner learn from this?
A marketing tag is code with full access to the page. An A/B testing tool, a tag manager, a chat widget and an analytics script can all read what a visitor types and change what a visitor sees. Whoever can edit those tools can edit your website, and on 16 November that turned out to include a stranger.
Four changes reduce the risk without giving up testing:
- Keep third-party scripts off checkout and login pages. Use your payment provider’s hosted fields, which the skimmer could not read. Where they apply, PCI DSS requirements 6.4.3 and 11.6.1 have been in force since 31 March 2025. They expect you to authorise every script on a payment page, check its integrity, keep an inventory with a reason for each, and be alerted to unauthorised changes to the page or its security headers.
- Write a CSP that names hosts. Report URI attributes the 24 blocked payloads to restrictive script policies. Restrict where pages can send data as well, and switch on reporting so a new destination raises an alert.
- Mark session cookies
HttpOnly. It costs nothing and puts them out of reach of any script on the page. - Give the tools an owner. Someone should know which platforms can inject code, who holds edit rights, and whether single sign-on and two-factor authentication are enforced on them.
This follows the same supply chain pattern as the poisoned Rank Math update in our round-up of October’s cyber incidents and the malicious plugins installed through UpdraftCentral. The attacker goes for the supplier, and the supplier’s customers deliver the malware for them.
What should visitors to affected sites do?
What to do depends on the device you used and whether you entered payment details or sent cryptocurrency:
- iPhone and iPad users should install the latest iOS or iPadOS update available for their device.
- Anyone who sent cryptocurrency on 16 November after copying an address on an affected site should compare the address the funds went to with the one they intended.
- Anyone who typed card details into an affected site that day should watch their statement and ask their bank about a replacement card.
What is still unknown?
Optimizely has issued an interim report and says its final root cause analysis will follow the investigation. Four questions stay open:
- When the first change was made. Report URI saw a poisoned BBC snippet at 04:21 UTC, and Optimizely dates the first malicious code written to a customer project to 07:51 UTC.
- How many projects were changed. Report URI identified 188 infected snippets linked to 178 organisations, and Optimizely says less than 5% of its customers were affected, without giving a number of customers or projects.
- How many card numbers, cookies and screenshots reached the attacker’s server, and whether any device was compromised. None of Report URI, Optimizely and Hudson Rock gives a figure.
- Whether one actor is behind both this attack and the related activity through Kissmetrics and StorifyMe scripts and the minute-ly video loader. Researchers have found shared loader code and hosting, which links the campaigns technically but does not prove a single operator. StorifyMe has confirmed that unauthorised JavaScript reached some of its customers.
The affected sites used Optimizely Web Experimentation, which gave its snippet permission to run on their pages. If you are not sure which scripts on your site could do the same damage, talk to AIWIZ and we will go through them with you.