🔍 Read the full analysis: Could Optimizely’s Website Malware Affect AI Users? Here’s What To Check on ThorstenMeyerAI.com
Get smart everyday buys delivered free with Prime
- Fast, free delivery on millions of items
- Prime Video, Amazon Music and more included
- Member-only deals all year
TL;DR
A security researcher says malicious JavaScript delivered through Optimizely’s experimentation platform reached visitors to sites associated with 178 organisations on Nov. 16, 2026. Optimizely says it fixed a vulnerability and removed the injected code; reported actions included wallet-address substitution, card-data collection and attempts to direct some Apple devices to exploit kits. Whether any visitors’ data or devices were actually compromised has not been established in the supplied reporting.
Optimizely confirmed that an unauthorized party exploited a vulnerability in its Experimentation platform to add malicious JavaScript to some customer projects, after security researcher Report URI said the code was served to visitors at 178 organisations on Nov. 16. The incident matters to people using AI tools on the web because an AI assistant or browser agent visiting an affected page could encounter the same injected code as any other visitor; the available reporting does not establish that AI systems or their users suffered a compromise.
Report URI said it found the malicious code in 188 of 1,611 Optimizely snippets it checked, and described the 178 organisations as those whose visitors received the code. The researcher cautioned that the actual number could be higher. BleepingComputer independently checked scans from urlscan.io and confirmed that the script loaded or attempted to load on sites including Zoom, KLM, GitLab, the BBC, CNN and Salesforce. A listed organisation does not mean its entire website or every visitor was affected: some entries were specific subdomains or tools.
The code was carried through Optimizely’s normal page-testing snippet, which customer websites had already chosen to load. Report URI said the attacker added code to the Project JavaScript field in customer projects, while the rest of each snippet remained unchanged. Optimizely told BleepingComputer that the vulnerability was in the Experimentation account signup flow. Its status page said it had fixed the vulnerability and removed the injected code from affected projects. The company also told BleepingComputer it had ended active sessions, revoked tokens and reset access for affected accounts.
Report URI’s analysis says the payload could replace cryptocurrency wallet addresses in page text, form fields and clipboard activity; monitor card details entered into a site’s own form fields; and collect cookies and screenshots. It said payment details entered in hosted payment frames, such as those used by Stripe or Adyen, were outside the script’s reach. The code also attempted to direct some iPhones and iPads to the Coruna and DarkSword exploit kits. Hudson Rock cautioned that its analysis showed delivery observations, not demonstrated iPhone compromises.
Optimizely script attack · Nov 13 – Nov 17, 2026
How one trusted script carried malware to 178 websites
The sites were not hacked themselves. The code arrived inside a file they loaded on purpose.
The route (schematic)
Whether any iPhone was compromised has not been shown.
Timeline (UTC)
- Nov 13Earlier wave through three other providers’ scriptsKissmetrics, StorifyMe, Minute Media
- Nov 16 04:21First Optimizely injection seen
- 08:28Address-swapping payload first seen on a site
- 13:43Card skimmer added
- 17:13iPhone exploit delivery widened
- 19:24Report URI sweep: 186 of 1,611 snippets infected
- 21:52Only 1 of 188 infected snippets left
- 23:11All infected Optimizely files cleaned
- Nov 17 01:02Optimizely confirms the incident on its status page
Sources: Report URI (Scott Helme), Nov 16, 2026, timeline and update of Nov 17; Optimizely status page, Nov 17, 2026; Hudson Rock (Infostealers.com), Nov 16–17, 2026. The route is a schematic.
Graphic: Powered by Thorsten Meyer AI
How the Script Could Reach AI Users
This incident shows how a compromise at a third-party web service can expose visitors without a direct breach of each site. The affected organisations had included Optimizely’s snippet in their pages for legitimate experimentation, so the malicious code arrived through a supplier those sites already trusted. Visitors generally would not be able to tell from the page alone that the script had changed.
That includes people using AI-powered browsers, assistants or agents to open websites, although the supplied sources do not report a specific AI product being targeted. If an AI tool loads a page in a browser environment, it may encounter scripts served to that page. The practical concern depends on what the tool or user does next: entering payment information into a site-controlled form or copying a wallet address while an affected tab is open could create exposure described by Report URI. This is a possible risk based on the reported script behaviour, not evidence that an AI service, account or device was compromised.
Report URI also found gaps in customer sites’ security policies: of 159 affected organisations it checked, 24 blocked the code and 68 had no policy. Those counts describe the researcher’s checks, not the security posture of every affected site. They show why controls that restrict which scripts a page may run can matter when a trusted supplier is compromised.
Products Worth Considering

Samsung 32-Inch Movingstyle Essential 4K UHD Smart Monitor with Rolling Stand

LG Smart Monitor Swing 32U889SA-W 32-Inch 4K UHD IPS Touchscreen on Fully Adjustable Wheeled Base, HDR10, webOS, Built-in Speakers, HDMI, US

FlexiSpot E6 Electric Standing Desk, 55 × 28 in, Bamboo Top, White Frame
As an affiliate, we earn on qualifying purchases.
How Optimizely’s Snippet Was Changed
Optimizely Web lets organisations run experiments on their pages using a small JavaScript snippet. Each customer’s snippet can load project-specific settings and optional custom code in a field called Project JavaScript, which can run across pages. Report URI said the attacker placed the malicious code in that field, so the affected pages continued to load a file from an address their operators already trusted. The researcher said the incident was not a CDN cache-poisoning attack.
Report URI recorded the first Optimizely injection it observed at 04:21 UTC on Nov. 16 and said all infected Optimizely files had been cleaned by 23:11 UTC that day, a roughly 19-hour observation window. The researcher said the payload was rebuilt 26 times during the day. These timestamps describe what Report URI observed; they do not prove that every customer or visitor was exposed for the full period.
Report URI also described an earlier wave beginning Nov. 13, in which the same actor’s injector appeared in scripts from Kissmetrics, StorifyMe and Minute Media. Those sites were separate from the published list of 178 organisations. Hudson Rock reported a similar pattern and said how the code entered those other services remained unresolved.
“An unauthorized party exploited a vulnerability in Experimentation and used it to add malicious JavaScript to some customer projects.”
— Optimizely, in its status-page update posted Nov. 17
The Scope of Exposure Is Unsettled
Neither the supplied statements nor the researcher’s findings establish how many people actually had data stolen, how many devices were compromised, or whether any particular visitor experienced financial loss. Report URI described the capabilities and observed delivery of the script; Hudson Rock explicitly said its iPhone findings did not demonstrate successful compromises. The presence of a company or subdomain on the list also does not establish that its full site, checkout system, customer base or every page was exposed.
The technical route from a flaw in the signup flow to write access across unrelated customer projects remains unexplained in the supplied material. Optimizely says it fixed the vulnerability and took steps to revoke access, but it has not publicly detailed the code-level mechanics in the cited reporting. The researcher’s 188-snippet check is also a sample, and Report URI said the total number may be higher. The sources do not say whether a visitor’s specific page loaded the code or whether a particular AI tool executed it.
What Customers and Visitors Can Check
People who visited a site named in Report URI’s list during the reported period can check with that organisation for incident-specific guidance, particularly if they entered card details into a form controlled by the site or copied a cryptocurrency address while the page was open. This is a precaution based on the code’s reported behaviour, not confirmation that those details were taken. Users should review relevant payment or wallet activity and contact the provider if they identify suspicious transactions.
For Apple-device users, Report URI advised updating to iOS 18.7.3 or later, or iOS 26.3 or later. The supplied source says those releases had been available for months by the time of the incident. Updating reduces exposure to known vulnerabilities but does not establish whether a device was affected in this event. Organisations using Optimizely should verify project access, review the company’s incident notices and check their own script restrictions and site logs.
Optimizely’s reported actions include fixing the signup-flow vulnerability, removing injected code, ending active sessions, revoking tokens and resetting affected account access. The next material information would be a fuller account of how the flaw enabled access across customer projects, the confirmed scope of affected projects and any findings about data theft or successful device compromise. The supplied reporting does not give a date for such disclosures.
Key Questions
Did Optimizely confirm a security incident?
Yes. Optimizely said an unauthorized party exploited a vulnerability in its Experimentation platform and added malicious JavaScript to some customer projects. It said it fixed the vulnerability and removed the injected code.
Were AI users or AI services confirmed to be compromised?
No AI-specific compromise is confirmed in the supplied reporting. AI-powered browsers or agents could encounter a script loaded by an affected page, but the sources do not establish that any AI product or user was compromised.
What could the malicious script do?
Report URI said it could replace cryptocurrency wallet addresses, monitor card details entered into sites’ own form fields, collect cookies and screenshots, and attempt to direct some Apple devices to exploit kits. Those reported capabilities do not establish that every action succeeded.
Does a company’s name on the affected list mean its whole website was compromised?
No. The list can refer to a particular subdomain or tool. Hudson Rock warned that an affected subdomain does not establish exposure across an organisation’s entire website, customer base or checkout system.
What should visitors do now?
Check for guidance from any affected organisation you visited, review payment or wallet activity if you entered details or copied an address, and install the latest software updates. These are precautions; the reporting does not confirm that a particular visitor’s information or device was compromised.
Source: ThorstenMeyerAI.com
