On 16 November 2026 malicious code was served to visitors of 178 organisations, including the BBC, Zoom, HSBC and Optimizely itself, through the Optimizely Web testing tool. None of those organisations was breached. The code arrived through third-party scripts they already loaded and already trusted.
That is the part worth absorbing. Your own servers can be perfectly secure and your website can still serve malware, because most sites execute code from a dozen companies they do not control and cannot inspect.
The attacker did not break Optimizely’s code. They wrote into the Project JavaScript field, the optional box where customers add their own scripts. Everything else in the snippet was untouched and it still came from Optimizely’s own CDN, so there was nothing to look suspicious.
What Actually Happened
The research is Scott Helme’s at Report URI, and it is worth reading rather than the coverage of it. He checked 1,611 Optimizely snippets and found 188 of them infected, belonging to 178 organisations, while noting the real number is probably higher because only the snippets he could find in time were checked.
The named list spans about twenty sectors and includes Zoom, the BBC, CNN, Salesforce, GitLab, IKEA, HSBC, Equifax, Air France-KLM, Ledger and Peacock. Optimizely’s own sites were among them, including its login page. None of those organisations was careless about security. They were loading third-party scripts, as every one of their competitors also does.
The timeline, in UTC on 16 November, is tight. The first injection appeared on a BBC snippet at 04:21. Zoom’s was modified at 11:20. Cryptocurrency was being swept to exchanges by 13:03 and a card skimmer was added at 13:43. Clean-up began at 20:41 and by 23:11 all the files Report URI monitored were clean.
Two things are not known. How the attacker obtained access has not been established. And Optimizely has acknowledged the incident on its status page, but Report URI reports receiving no response to its outreach, so treat any account of the vendor’s internal findings carefully.
Why No Server Was Breached
Because nothing needed to be. This is the mechanism that makes third-party scripts a category of risk on their own, rather than a variation on ordinary intrusion.
A tag like Optimizely’s works by having you place a small script on every page. That script is fetched from the vendor’s CDN and runs with the same privileges as code you wrote yourself. It can read the page, read what users type, and change what they see.
The attacker used the Project JavaScript field, the part of the product designed for customers to add their own code. So the file served to the BBC’s visitors was the BBC’s genuine Optimizely snippet, from Optimizely’s genuine CDN, with extra instructions inside it. Nothing about the request looked wrong, because nothing about it was wrong.
This is why reaching for the usual tools does not help. There was no vulnerable server to patch, no suspicious domain to block, no phishing email to catch. The defence has to sit somewhere else entirely.
Not sure what runs on your site?
Talk to the Transputec team about the security work behind our Managed SOC and Vulnerability Management.
Get in TouchWhat the Code Was Doing
Four things, and only two of them made the headlines.
| Behaviour | Who it hurts |
|---|---|
| Swapping cryptocurrency wallet addresses | Anyone copying an address on an affected page. Roughly 7,040 US dollars had reached the attacker’s wallets by 19:25 UTC |
| Skimming payment cards from forms | Customers of affected sites, from 13:43 UTC onwards |
| Stealing cookies and screenshots | Logged-in users, including staff on internal-facing pages |
| Serving iOS exploit kits | iPhone and iPad visitors, via the Coruna and DarkSword kits |
The third row is the one most readers skip and the one most relevant to a business audience. Cookie theft from an authenticated session is a route into an account that never touches a password or a second factor. If your staff use a browser on a site carrying third-party scripts, and almost everyone does, that is a path worth understanding.
Why You Cannot See Inside a Third-Party Script
Because the file can change after you approve it, and usually does.
A typical business website loads code from analytics, tag management, chat widgets, A/B testing, advertising pixels, consent banners, fonts, video players and payment providers. Each one is a company whose engineering decisions become your security posture, and almost none of it was reviewed after the day it was added.
Three properties make third-party scripts harder to govern than ordinary supplier risk.
- The content is not fixed. You approved a URL, not a file. What that URL returns is the vendor’s decision, and it can differ tomorrow or differ by visitor.
- Scripts load other scripts. A tag manager is a loader. What it pulls in is frequently invisible to whoever signed the contract.
- It executes on the user, not on you. Your logs may show nothing at all, because the code ran in a visitor’s browser, not in your estate.
That last point is why detection has to happen at the browser rather than at the perimeter, and why Report URI found this at all: it monitors what browsers actually receive.
What to Do About It
Five things, in order of how much they cost you.
- Write the inventory. List every third-party script on your site, what it does, who owns it internally and whether it is still needed. Most organisations find things nobody can account for, often from campaigns that ended years ago.
- Delete what you are not using. The cheapest security work available. A tag nobody reads data from is pure exposure.
- Set a Content Security Policy. It tells the browser which origins may execute code. It would not have stopped this attack, because the code came from an origin you trust, but it stops the much larger class of attacks where it does not.
- Monitor what browsers actually receive. Reporting on policy violations is how a change like this surfaces, and it is how this one surfaced.
- Keep sensitive flows away from shared pages. A payment or login page carrying a dozen marketing tags is a choice, and it is worth revisiting.
The NCSC’s supply chain security guidance covers the wider discipline. The first two items above need no budget and no supplier, which is usually where to start.
The Reporting Angle for Regulated Firms
If you are FCA regulated, an incident of this shape is about to come with paperwork.
The FCA’s PS26/2 on operational incident and third party reporting was published on 18 March 2026 and comes into force on 18 March 2027. It requires reporting of qualifying operational incidents through a single route, and a register of material third party arrangements submitted to the FCA annually. In setting the rules the FCA states plainly that many of the incidents reported to it originate at third parties.
A marketing testing tool is not obviously a material third party. Then it serves a card skimmer to your customers, and the question of whether it should have been on the register answers itself. Three things are worth settling now rather than in March 2027.
- Does your register include the vendors whose code executes on your public site, or only the ones holding your data?
- What is your reporting threshold when the failure is at a supplier and your own systems are untouched?
- Who decides, and could you produce a timestamped timeline like the one above for your own estate?
What to Ask Anyone Whose Code Runs on Your Site
Six questions. Ask them of the vendors you already have, not only the next one.
- Who at your end can change the code served to our visitors, and how is that access controlled? This is the question this incident turns on.
- Is multi-factor authentication enforced on those accounts?
- Do you log changes to customer-editable fields, and will you give us those logs?
- How quickly would you tell us, and through what channel? A status page is not a notification.
- Can we pin a version, or do we always receive your latest?
- What is your own supply chain? Their script may load somebody else’s.
The fourth is the one that bites. Several affected organisations learned about this from a security researcher’s blog rather than from their vendor, which is a notification path nobody designed and nobody should rely on.
If You Think You Were Affected
Treat it as a customer data incident rather than a website problem, because that is what it is.
- Establish the window. When did your snippet change and when was it cleaned? Your vendor should be able to tell you; Report URI’s timeline gives you a reference to check it against.
- Work out what was exposed in that window. Card details entered on affected pages, session cookies for logged-in users, anything typed into a form.
- Invalidate sessions. Stolen cookies do not expire because the script was removed.
- Consider your notification duties. If personal data was accessible, UK GDPR timelines apply regardless of whose code caused it.
- Keep the evidence. The snippet as served, the timestamps, what you did and when.
The uncomfortable part is that most organisations cannot do the second step, because nobody records what third-party scripts were serving at a given hour. That gap is the real finding from this incident, and it is worth closing before the next one.
How Transputec Approaches Third-Party Scripts
Honestly, this is a gap most organisations have and few have measured. The work is unglamorous and it is mostly inventory before it is ever tooling.
Transputec runs six cyber security services from a UK SOC, and two of them matter here. Managed SOC covers the hours when a change like this one lands at 04:21 and nobody is looking. Vulnerability Management is where the external attack surface gets reviewed rather than assumed, and a script inventory belongs in that review alongside the usual open ports and certificates.
Where a vendor’s code is the exposure, our guide to assessing third-party data security covers the procurement half of the question.
If you cannot currently list the third-party scripts running on your own website, that is the first conversation, and it is a short one. Talk to our team about the security work behind our Managed SOC.
Frequently Asked Questions
What happened in the Optimizely attack?
On 16 November 2026 an attacker added malicious code to the Project JavaScript field of customer projects in Optimizely Web, the optional area where customers place their own scripts. Report URI checked 1,611 Optimizely snippets and found 188 infected, belonging to 178 organisations including the BBC, Zoom, HSBC, CNN, Salesforce, GitLab, IKEA and Optimizely’s own sites. The code swapped cryptocurrency wallet addresses, skimmed payment cards, stole cookies and screenshots, and served iOS exploit kits. How the attacker gained access has not been established.
Were the affected websites hacked?
No, and that is the significant part. None of the 178 organisations had its own servers compromised. The malicious code was delivered through a script those sites already loaded from Optimizely’s CDN, served from a legitimate URL inside a legitimate file. There was no vulnerable server to patch and no suspicious domain to block, because the request was genuine in every respect except what the file contained.
What are third-party scripts and why are they risky?
Third-party scripts are code your website loads from another company: analytics, tag managers, chat widgets, A/B testing, advertising pixels, consent banners, video players and payment tools. They run with the same privileges as your own code, so they can read the page, read what users type and change what visitors see. Three things make them a distinct risk. You approved a URL rather than a file, so the content can change after review. Scripts can load other scripts, so the chain is often invisible. And the code executes in your visitor’s browser rather than in your estate, so your own logs may show nothing.
Would a Content Security Policy have stopped this?
Not this one. A Content Security Policy tells the browser which origins are allowed to execute code, and here the malicious code came from an origin the sites had deliberately trusted. It is still worth having, because it blocks the much larger class of attacks where code is injected from somewhere unexpected, and because the reporting it generates is how changes like this become visible at all. The lesson is that CSP is a floor rather than a complete answer, and it needs to be paired with monitoring what browsers actually receive.
Do we have to report a third-party incident to the FCA?
Possibly, and from 18 March 2027 the position is firmer. The FCA’s PS26/2, published 18 March 2026, requires qualifying operational incidents to be reported through a single route and a register of material third party arrangements to be submitted annually. The FCA notes that many incidents reported to it originate at third parties, so an incident you did not cause can still be one you must report. Two practical questions follow: whether your register covers vendors whose code executes on your public website rather than only those holding your data, and what your reporting threshold is when your own systems are untouched.
This article was drafted with AI assistance and reviewed by the Transputec team. Featured image: AI-generated.



