← Back to Blog
Hot TopicOn This Site

The Optimizely Snippet Hack (OPTIMJACK, November 2026): What Happened to 178 Websites, and How to Lock Down Third-Party Scripts on Yours

On 16 November 2026 an attacker abused a flaw in Optimizely Web Experimentation's signup flow to plant malicious JavaScript in customer projects. Researchers counted 188 infected snippets across 178 organisations, serving a crypto address swapper, a card skimmer and iOS exploit kits for most of a day. Only about 15 percent of the sites they checked had a Content Security Policy that stopped it. Here is what Optimizely confirmed, what to check if you use it, and the third-party script lockdown I'd do at any startup this week.

·10 min read

On 16 November 2026, an attacker exploited a flaw in the account signup flow of Optimizely Web Experimentation, the A/B testing tool, and used it to add malicious JavaScript to some customer projects. Because those projects are delivered as a script tag on the customer's own pages, the code ran on the live websites of banks, media companies and SaaS brands for most of a day. Researchers at Report URI counted 188 infected snippets belonging to 178 organisations, and the payload swapped crypto wallet addresses, skimmed card details from on-page forms and served iOS exploit kits. None of those companies had their own servers breached. If you use Optimizely Web Experimentation, check your project code history and talk to Optimizely today. If you don't, the lesson still applies to you: every third-party script on your site can rewrite your checkout page, and most sites have nothing in place to stop it. Below is what is confirmed, and the script lockdown I'd do this week.

Who should care

  • Optimizely Web Experimentation customers. Optimizely says it has contacted every Web Experimentation customer directly to confirm whether they were affected. If you haven't heard, chase it.
  • Anyone with marketing, analytics or chat scripts on checkout, login or account pages. A/B testing tools, tag managers, heatmaps, chat widgets and review widgets all run with full access to the page.
  • Startups that take payments with on-page card forms instead of a hosted payment field or redirect.
  • Teams without a Content Security Policy, or with one that allows unsafe-eval and any host.
  • Visitors of affected sites (crypto users especially), who should check wallet transactions and card statements from 16 November.

If your product has no third-party JavaScript at all, you are in a small minority. Everyone else should keep reading.

What actually happened

Optimizely's own account

Optimizely's status page carries the incident as "Web Experimentation Security Incident", affecting snippet delivery. Its first update, posted 17 November 2026 at 01:02 UTC, says an unauthorized party "exploited a vulnerability in Experimentation" to add malicious JavaScript to some customer projects, and that Optimizely had fixed the vulnerability and removed the injected code.

The second update, posted 18 November at 14:49 UTC, is more specific: on 16 November, an unauthorized third party exploited a flaw in the Web Experimentation account signup flow. Optimizely says this "was not a credentials breach", that the incident was limited to Web Experimentation, that the flaw was closed within three hours of detection, that less than 5 percent of its customers were impacted, and that no malicious code remains in its systems. Coverage by TechTarget and Interesting Engineering adds that Optimizely ended active sessions, revoked authentication tokens and reset access for affected accounts.

What Optimizely has not explained is how a signup flaw gave the attacker write access to many unrelated customers' projects. The Cloud Security Alliance's research note calls that a gap and infers some form of cross-tenant write access; that is its inference, not a confirmed root cause. There is no CVE. When I checked the status page for this post, the incident was still listed as unresolved, and Optimizely says it is working with third-party experts on a full investigation.

What researchers saw

Most of the detail comes from Report URI (Scott Helme's team), relayed by Help Net Security, TechTarget, CSA and others:

  • Where the code went. Into the customer-editable Project JavaScript field, which Optimizely then shipped inside each customer's snippet. This was not CDN cache poisoning.
  • Scale. Report URI checked 1,611 snippets and found 188 infected, belonging to 178 organisations; it says the real number is likely higher. Financial companies were the largest group, with 32 organisations. Named victims in coverage include the BBC, CNN, Zoom, Salesforce and HSBC.
  • Timeline (UTC, per Report URI as summarised by CSA and others). First injection around 04:21 on 16 November; a card skimmer was added around 13:43; mass cleanup ran from about 20:40, and by 21:52 only one infected snippet remained. Estimates of the exposure window range from about 17 to about 19 hours depending on who is counting.
  • Payload. A multi-purpose kit, rebuilt 26 times during the day: it swapped crypto wallet addresses, skimmed card numbers, expiry dates and CVVs from on-page forms, collected cookies that weren't marked HttpOnly and screenshots, and loaded the Coruna and DarkSword iOS exploit kits for Apple devices. Sources describe the iOS version targeting differently, and Hudson Rock has cautioned that scanner results are not proof of compromised phones.
  • Money. Wallets linked to the attacker received about $7,000 (Report URI said about $7,040 on the evening of 16 November; Ostorlab later put it at about $7,150), and CSA notes some of those wallets were active before the attack, so not all of it came from victims. Nobody can say how many cards were skimmed.
  • Earlier wave. Help Net reports tampered responses from other third-party services (Kissmetrics, StorifyMe and minute-ly; CSA names the last one Minute Media) from 13 November. Treat that as a separate thread still being investigated.

The name OPTIMJACK came from independent researchers, not from Optimizely, according to CSA.

What the affected companies said

Per Reuters reporting summarised by HackerNoon: Salesforce said it was investigating and disabling scripts as a precaution; Zoom said it disabled Optimizely in production and found no customer data impact at that point; Equifax said it stopped using Optimizely and found no evidence its own systems were compromised. None of this proves visitors were told.

The part that should worry every startup

Report URI also classified the Content Security Policies of 159 affected organisations. 68 had no CSP at all. 50 had one that let the malicious script load and send data out. 17 blocked the data from leaving but not the script. Only 24, about 15 percent, blocked the payload.

Two details stand out. At HSBC, per Helme, the CSP blocked the external payload, but the attacker's inline fallback ran anyway because the policy allowed unsafe-eval. And some sites that proxied or self-hosted the Optimizely file were still infected, because they were faithfully copying the poisoned file. Helme put the core lesson plainly in TechTarget: "none of the victims were breached."

Every script tag on your site is a vendor with write access to your checkout page. Treat it like one.

If you use Optimizely Web Experimentation: what to check

  1. Get Optimizely's answer in writing. Ask whether your account was affected and request the "additional guidance" its status page mentions. Keep the reply for your incident file.
  2. Review Project JavaScript and change history for 13 to 17 November. Look for script tags, atob(, eval( or new code you didn't write.
  3. Search your own logs and CSP reports for the known indicators back to 04:00 UTC on 16 November: the payload host 1688360.cc (/track.js), 5vobauua3ojpklx.icu, and the global window._k7. Indicator lists from vendors differ; treat these as a starting point.
  4. Review users and tokens. Remove accounts you don't recognise, rotate API tokens, enforce SSO and MFA, and cut editor rights to the few people who actually publish experiments.
  5. Decide on card and session exposure. If an affected snippet ran on a page with an on-page card form during the window, Unit 42's advice is to treat card data entered there as exposed. Talk to your payment processor and your lawyer about notification duties (PCI DSS, GDPR, India's DPDP Act). Invalidate sessions held in cookies readable by JavaScript.

A quick way to see whether a snippet is clean right now (replace the ID with yours; it only tells you about today, not 16 November):

curl -s https://cdn.optimizely.com/js/YOUR_PROJECT_ID.js \
  | grep -Eo '1688360\.cc|_k7|atob\(|eval\(' | sort | uniq -c

The third-party script lockdown (for everyone)

This is the practical core, and it applies whether or not you have ever heard of Optimizely.

1. Inventory what actually runs

You can't defend scripts you don't know about. Tag managers and widgets load other scripts, so read the page as a browser sees it, not your codebase. This Playwright script lists every third-party script host on your key pages:

// list-third-party-scripts.mjs  (run: node list-third-party-scripts.mjs)
import { chromium } from "playwright";

const site = "https://example.com";
const pages = ["/", "/pricing", "/login", "/checkout"];
const browser = await chromium.launch();
const page = await browser.newPage();
const hosts = new Map();

page.on("request", (req) => {
  if (req.resourceType() !== "script") return;
  const host = new URL(req.url()).host;
  if (!site.includes(host)) hosts.set(host, (hosts.get(host) || 0) + 1);
});

for (const p of pages) await page.goto(site + p, { waitUntil: "networkidle" });
console.log([...hosts.entries()].sort((a, b) => b[1] - a[1]));
await browser.close();

For each host, write down the owner, why it's there and which pages need it. Anything nobody can justify gets removed.

2. Get scripts off sensitive pages

TechTarget quotes Helme on this: an A/B testing tool rarely needs to be on a checkout or account page at all. I'd go further. Login, signup, checkout, billing, password reset and admin pages should load your own code and your payment provider, and nothing else. In Next.js that usually means putting analytics and widgets in a layout that the sensitive routes don't use, rather than in the root layout.

3. Never let card numbers touch your DOM

Use your payment provider's hosted fields or hosted checkout (Stripe Checkout or Elements, Adyen's hosted components, Razorpay Checkout). ThreatDefence notes that hosted payment frames from Stripe and Adyen were out of the skimmer's reach. A script on your page can read any input on your page; it cannot read inside a cross-origin iframe.

4. Ship a Content Security Policy, in report-only mode first

A CSP lists which hosts may run scripts, which hosts the page may send data to, and where forms may post. Start in report-only mode so nothing breaks, read the reports for a week, then enforce. For a Next.js app, a starting point in next.config.ts:

const csp = [
  "default-src 'self'",
  "script-src 'self' https://www.googletagmanager.com",
  "connect-src 'self' https://www.google-analytics.com",
  "img-src 'self' data: https:",
  "style-src 'self' 'unsafe-inline'",
  "form-action 'self'",
  "frame-ancestors 'none'",
  "base-uri 'self'",
  "object-src 'none'",
].join("; ");

// inside headers(): add to the headers array for source "/(.*)"
{ key: "Content-Security-Policy-Report-Only", value: csp },

The rules I'd hold to: no unsafe-eval (that is what let the HSBC fallback run); a tight connect-src, because that is the line that stops stolen data leaving (17 organisations were saved by it alone); form-action 'self'; and a strict policy on sensitive pages even if marketing pages need a looser one. Next.js documents a nonce-based CSP through middleware if you need to drop unsafe-inline for scripts; it costs you static rendering on those routes, so decide per page.

5. Know what SRI can and can't do

Subresource Integrity pins a script to a hash, so a changed file won't run. It is excellent for versioned libraries from a CDN. It does not work for snippets like Optimizely's that change every time someone publishes an experiment, and self-hosting a copy you re-sync from the vendor just copies the poison, as some victims found.

6. Treat vendor consoles as production access

Whoever can edit your tag manager, A/B tool or chat widget can run code on your checkout. Put those tools behind SSO with MFA, give publish rights to as few people as possible, review the user list quarterly and turn on change notifications where the vendor offers them. This is the same vendor-access discipline I described for cloud roles in the Expedia and Qubole post.

7. Make cookies boring to steal

Session cookies should be HttpOnly, Secure and SameSite=Lax or stricter. The payload collected cookies readable from JavaScript; an HttpOnly session cookie isn't one of them.

8. Watch for change

CSP violation reports, sent to an endpoint you or a monitoring service collects, are your early warning: the first time a new host appears on your checkout, you want to know the same day. Several affected companies learned about this from researchers and the press.

What I'd do as a CTO this week

Monday: run the inventory on the five pages that matter most and kill anything without an owner. Tuesday: move card entry to hosted fields if it isn't already, and pull every non-essential script off login and checkout. Wednesday: ship a report-only CSP. Thursday: put SSO, MFA and a short editor list on every tool that can inject code. Following week: enforce the CSP on sensitive routes once the reports are quiet.

None of this is expensive. It is mostly deciding that marketing tooling doesn't get to run wherever it likes. That is a conversation between growth and engineering, and in early startups it usually doesn't happen until something like this forces it. It sits alongside the basics in my startup security checklist and fits the same supply-chain pattern as the wfx-loot GitHub Actions stealer: the code you trust most is the code attackers want to sit inside. If nobody on your team owns this, it is exactly the kind of gap a fractional CTO or a technical advisor closes in the first month.

FAQ

What is OPTIMJACK?

It is the name independent researchers gave to the 16 November 2026 campaign in which an attacker used a flaw in Optimizely Web Experimentation's signup flow to plant malicious JavaScript in customer projects. The code ran on customers' websites and stole crypto, card details from on-page forms, cookies and screenshots, and served iOS exploit kits.

Which websites were affected by the Optimizely hack?

Report URI found 188 infected snippets belonging to 178 organisations, with financial companies the largest group; names in coverage include the BBC, CNN, Zoom, Salesforce and HSBC. Optimizely says less than 5 percent of its customers were impacted and has contacted every Web Experimentation customer directly.

Is Optimizely safe to use now?

Optimizely says it closed the flaw within three hours of detection, removed the injected code, revoked tokens and reset access, and that no malicious code remains. Its investigation was still open when I wrote this. Whatever tool you use, keep it off sensitive pages and behind a CSP, so the next vendor incident can't reach your checkout.

I visited an affected site on 16 November. What should I do?

Check card statements for unfamiliar charges and verify recent crypto transfers went to the address you intended. If you use an older iPhone, update iOS. Keep an eye out for notices from the site itself.

Would a Content Security Policy have stopped it?

For about 15 percent of the sites Report URI checked, yes. A CSP that only allows known script hosts, bans unsafe-eval and restricts where data can be sent would have blocked the payload or the theft. Most sites had no CSP or one too loose to matter.

Does Subresource Integrity (SRI) protect against this?

Not for snippets like Optimizely's, which change whenever an experiment is published, so there is no fixed hash to pin. SRI is still worth using for versioned libraries loaded from a CDN.

Want a second pair of eyes on what your site actually loads, and whether your checkout is one vendor incident away from a skimmer? Book a 30-minute call and we'll go through it.

Sources

Written By

Kunal Vohra

Kunal Vohra

Technical Co-Founder & Fractional CTO

I've co-founded 6+ startups across India, the UAE, and the US, spanning AI, Web3, fintech, and cybersecurity. I write about the technical and strategic decisions that determine whether a startup thrives or stalls.

Comments

Loading comments…

Enjoyed this article?

More writing on AI, Web3, and building startups.