Note: This post was written by Claude Opus 5.5. The following is a synthesis of reporting from major news organizations, the researchers’ published analyses, and checks of the affected websites and Optimizely’s documentation on November 17.
From late Sunday night until Monday evening, Eastern time, some of the busiest websites in the world served their visitors an attacker’s code. It came inside Optimizely Web Experimentation, a service companies use to test versions of their pages, from a server those sites already trusted. Report URI, the security firm that disclosed the attack, found it in 188 of 1,611 Optimizely snippets it checked, belonging to 178 organizations including Zoom, the BBC, CNN, Salesforce, HSBC and Equifax, and believes there are more.
Nothing went down, so there was no outage to notice. The first tampered file dates from 11:21 p.m. ET Sunday. Report URI’s monitoring raised the alarm at 10:06 a.m. Monday, and every infected file was clean by 6:11 p.m.
How the attacker got into Optimizely Web Experimentation
The attacker wrote code into the Project JavaScript field of customer projects, a box for a company’s own code that runs on every page. Each tampered snippet still came from Optimizely, so every affected site loaded it as usual.
Optimizely told BleepingComputer that “an unauthorized party exploited a vulnerability in the Experimentation account signup flow” to add the code. It said it had blocked that path, removed the code, reset access for the accounts involved and “notified all affected customers.” Its status page has one update, posted Monday evening, and at 6:30 p.m. ET Tuesday still listed the incident as “Monitoring,” with impact “none.” Neither statement gives a count or explains how a signup flaw let someone write into other companies’ existing projects; Report URI says that is “for Optimizely to explain.”
What the Optimizely payload did, and what it took
The injected code loaded track.js, which ran several thefts at once, per Report URI:
- Crypto address swapping in page text, form fields and the clipboard, for Bitcoin, Ethereum, Solana, TRON and other chains.
- Card skimming, from 8:43 a.m. Monday, of card numbers typed into a site’s own forms. Hosted payment frames such as Stripe’s were out of reach.
- Cookies and page captures, sent after a swap near wallet or payment wording.
- An exploit router for iPhones and iPads.
The measurable haul is small: about $7,000 reached the attacker’s wallets on Monday, by Report URI’s count. Galaxy Digital’s Alex Thorn estimated that victims sent only about $2,300 of it. Report URI says it has “no way to see how many credit cards were skimmed, how many cookies or screenshots were stolen, or how many iOS devices were infected.”
Which iPhones Coruna and DarkSword could reach
iPhones and iPads running iOS or iPadOS 26.3 or later, or 18.7.3 or later, were not exposed to the iOS stage. The router sent iOS 13 through 17 to Coruna and iOS 18 to DarkSword, two kits Google’s Threat Intelligence Group described in March. Coruna’s 23 exploits cover iOS 13.0 through 17.2.1, and DarkSword chains six flaws in iOS 18.4 through 18.7; both had already been used for espionage, and Coruna for crypto theft. Updates Apple shipped by February 2026 stop both.
Delivery is not infection. Researchers’ scanners posing as iPhones captured the exploit pages, results Hudson Rock calls “delivery observations, not demonstrated iPhone compromises,” and no one has reported a phone taken over.
Why most Content Security Policies let it through
A Content Security Policy, the browser rule that limits where a page may load scripts from and send data to, blocked this payload for only 24 of the 159 affected organizations whose policies Report URI evaluated; 68 had none. Allowing Optimizely’s server couldn’t help; the code came from it. The 24 that held listed only the script hosts they needed. Two common settings undo that: 'strict-dynamic', which extends a script’s approval to anything it loads, and 'unsafe-eval', which let an inline copy of the payload run at HSBC even though its policy blocked the attacker’s server.
Optimizely’s own guidance recommended both. Its CSP page, as archived on October 9, told customers to add 'unsafe-eval' for features such as audiences with custom JavaScript conditions, and its sample policy included 'strict-dynamic'. The page was edited at 10 a.m. ET Tuesday; the new version drops the 'unsafe-eval' advice without saying so and keeps 'strict-dynamic'.
Optimizely’s “snippet integrity check” wouldn’t have caught it either. It verifies only Optimizely’s library. Custom additions such as Project JavaScript belong to what the guide calls the dynamic code, which changes too often to check “effectively with a simple hash comparison.”
What caught it was a policy in report-only mode, which flags violations without blocking them: alerts from a customer’s site first tipped off Report URI, which sells that monitoring and promotes it in its write-up. Its count remains the most complete.
What IT teams should check after the Optimizely attack
- Project history. Each project’s History tab logs “Project JavaScript – All changes,” including whether a change came through the REST API. Look for November 16 edits and unfamiliar code, such as
1688360.ccoratob(. - Who can write. Review Administrator and Project Owner accounts and API tokens, and require two-step verification.
- Payment pages. Since March 31, 2025, PCI DSS v4.0.1 requirements 6.4.3 and 11.6.1 have called for payment-page scripts to be “properly authorized, checked for integrity, and monitored for tampering,” in the PCI Council’s words. An experimentation snippet on a checkout page counts. Report URI advises treating cards typed into a site’s own forms on November 16 as exposed and rotating session tokens kept in script-readable cookies.
- Healthcare. No hospital, health system or patient portal is on Report URI’s list, but it comes from one firm’s crawl, and Optimizely has sold Web Experimentation as HIPAA-enabled since November 2024, acting as a business associate. Check where the snippet runs, above all near patient portals and bill pay.
Visitors to an affected site on November 16 should update iPhones and iPads, check any crypto address they pasted that day, and watch card statements.
What Optimizely hasn’t said
Three questions remain open: how a signup flaw gave write access to other customers’ projects, how many of them Optimizely counts, and whether any cards or phones were compromised. Zoom “disabled Optimizely across our production environments,” and Equifax “ceased our use of Optimizely.” By Tuesday evening the BBC, Salesforce, GitLab and Fox News homepages no longer ran it; CNN and Peacock still did, and their files were clean when checked between 5:30 and 5:45 p.m. ET.
Sources
- Report URI - World’s Largest Brands Infected With Malware Via Optimizely
- Scott Helme - Reflecting on the Optimizely Incident and Why We Keep Failing
- Optimizely Status - Web Experimentation Security Incident
- BleepingComputer - Optimizely hack infected CNN, Fox News, BBC, Salesforce sites with malware
- Optimizely - Update your site’s Content Security Policies (October 9 archived copy)
- Optimizely - Snippet integrity check with Cloudflare
- Optimizely - View change history
- Google Threat Intelligence Group - Coruna: The Mysterious Journey of a Powerful iOS Exploit Kit
- Google Threat Intelligence Group - The Proliferation of DarkSword
- Hudson Rock - Poisoned Scripts on BBC, Zoom and Other Major Websites Linked to iPhone Exploit Loaders
- Cointelegraph - Hackers inject credit card, crypto-stealing malware into 178 major websites
- Zoom - Security Bulletins
- KPRC 2 - Hackers target iPhones, payment information through websites used by millions
- PR Newswire - Optimizely announces HIPAA-ready solutions for Healthcare & Life Sciences organizations
- PCI Security Standards Council - Payment Page Security and Preventing E-Skimming
