On 16 November 2026 an attacker wrote malicious code into the Optimizely snippets of 178 organisations, and Optimizely’s own content delivery network served it to their visitors. Nobody’s website was hacked in the usual sense. The code arrived through a script tag every one of those sites had deliberately added, and it ran with full access to every page. Scott Helme of Report URI published the timeline and the analysis the same day, and everything below is drawn from it. Report URI, 16 November 2026.
This post is the record of the incident. The evergreen question, how these attacks work and how to stop your site being next, is in our guide to web supply chain attacks and third party scripts, and the website script check will tell you in a few seconds what the payload would have done on your own site.
The timeline
Optimizely is an experimentation and personalisation platform. Customers add one script tag, served from Optimizely’s CDN, and manage what it does from the Optimizely console. The attacker wrote code into the project configuration of 188 snippets belonging to 178 organisations, among them the BBC, Zoom, Salesforce and IKEA, so the CDN served exactly what it was given.
- 04:21 UTC. The first injection is seen, on the BBC’s snippet.
- 13:43 UTC. Card skimming is added to the payload. It was rebuilt 26 times during the day as the attacker added capabilities.
- 19:24 to 19:59 UTC. Report URI’s sweep finds 186 of 1,611 snippets infected.
- 20:40 to 21:10 UTC. The mass clean up runs.
- 23:11 UTC. Every infected snippet is reported clean. Optimizely acknowledged the incident on its status page.
The same injector had appeared in three other vendors’ scripts between 13 and 16 November, so this was a campaign against the web supply chain rather than one bad day for one company. How the attacker gained write access to customer projects had not been published at the time of writing. Report URI lists a compromised Optimizely system, a compromised credential with access across customers, or something else, and says the true number of affected sites is likely higher than 178.
What the payload did to visitors
Four things, in the order they were added.
- Cryptocurrency theft. Wallet addresses were replaced in page text, forms, the clipboard, outgoing requests and the iOS share sheet. About $7,000 was taken on the first day, and the funds were swept out at 13:03 UTC.
- Card skimming. Card number, expiry and CVV were captured from forms on the page and posted to an attacker domain. Cards entered inside a hosted payment frame were out of reach, which is a point in favour of hosted payment pages.
- Session and screen theft. Cookies not marked HttpOnly, the page URL and a rendered screenshot of the page were sent to the same attacker domain.
- Device exploitation. A hidden frame loaded iOS exploit kits: Coruna for iOS 13 to 17, DarkSword for iOS 18, and a fingerprinting chain for iOS 26. On older versions the page reloaded every 60 seconds to retry. Devices on iOS 18.7.3 or 26.3 and later were not affected.
The last one is the part that matters for a corporate site with no payments and no customer accounts. Any script on your page can open a frame, and an unpatched phone visiting your domain is compromised by a page your domain served. The damage lands on the visitor, and the incident report names you.
What actually worked, in numbers
Report URI crawled the affected organisations’ sites and grouped them by what their Content Security Policy did when the payload arrived.
| Outcome | Sites | What it means |
|---|---|---|
| No policy at all | 68 | Payload ran, data left the page, exploit frames loaded. |
| Policy allowed both | 50 | A policy existed but permitted the vendor script and the attacker’s hosts. No better than none. |
| Payload ran, theft blocked | 17 | A restrictive connect-src stopped cards, cookies and screenshots leaving the page. Address swapping, which happens inside the page, still worked. |
| Payload blocked | 24 | A script-src that listed only the hosts actually needed. The malicious code never ran. 15 percent of the affected sites. |
Three findings are worth repeating to whoever owns your website. Allowing the vendor’s CDN in the policy did not help, because the vendor’s CDN is where the payload came from. Two common shortcuts defeated the policy: strict-dynamic, which lets an allowed script load whatever it likes, and unsafe-eval, which let one site’s inline fallback run the payload after its tag was refused. And a tight connect-src is a second wall that costs nothing, because it stops the theft that depends on sending data out even when the script runs.
What to check on your own site this week
The website script check reads one page and tells you which vendors it loads script from, what its tag manager container may load, and whether its policy would have blocked, limited or allowed each step of this payload. Run it on the homepage and then on the contact page. If the answer is “no policy” or “allowed”, you are in the group of 118 organisations whose sites ran the payload in full on 16 November, and the fix is a few days of configuration by someone who understands the policy.
The guide covers the mechanism, the earlier incidents that worked the same way, and the defence in order. If you would rather we did the read, send us the domain and we will go through the whole site, free and in writing: talk to us.