Hackers exploited a flaw in Optimizely’s web-testing software on November 16, 2026, exposing visitors to 178 major websites to crypto theft via malware and payment-card skimming. The affected organizations included HSBC, Ledger, Salesforce and the BBC, but attackers reached their websites through a shared vendor rather than breaching each site directly.
Key takeaways
- A flaw in Optimizely’s signup process let attackers alter customer scripts.
- The malware swapped wallet addresses across nine cryptocurrency networks.
- Optimizely blocked the vulnerable route, deployed a fix and notified affected customers.
One software flaw reached 178 organizations
Security researcher Scott Helme and his team identified 188 infected Optimizely JavaScript files associated with 178 organizations after examining 1,611 scripts. Optimizely’s software lets businesses test page layouts and wording; visitors’ browsers automatically run its embedded code when loading those pages.
The affected organizations spanned finance, technology, media and other sectors. Alongside HSBC, hardware wallet maker Ledger, Salesforce and the BBC, affected websites included Zoom, GitLab, KLM, Air France, Bitly and UNICEF USA.
Among the entities identified in Helme’s research, six were companies operating in crypto or trading: Ledger, IG Group Holdings Plc, OANDA’s Coinpass, Interactive Investor, TMS Brokers and Betashares, while 32 others were banks and financial companies.
Crypto theft via malware and payment-card skimming
The malicious code searched webpages for cryptocurrency wallet addresses and substituted destinations controlled by the attackers. It targeted Bitcoin, Ethereum, Solana, Tron and five other networks, changing addresses in form fields and in text copied to visitors’ clipboards.
A visitor who pasted a substituted address into a wallet and authorized the transfer would send funds to the attacker. This method of crypto theft via malware did not require access to the victim’s wallet: it changed the destination before the transaction.
The script also captured payment-card numbers, expiration dates and security codes entered on webpages. Its other capabilities included collecting browser cookies, webpage addresses and screenshots, and delivering exploits targeting older iPhones and iPads. Helme identified the secondary payloads as the Coruna and DarkSword exploit kits.
The injected JavaScript loaded an additional malicious script called track.js. Some versions encoded its address, while Helme found another variation with an embedded payload that could execute if the remote script was blocked.
Optimizely closes the signup vulnerability
Optimizely said an unauthorized party exploited its Experimentation account signup flow to place malicious JavaScript in customer projects. That vendor-level entry point allowed the code to reach customer websites without separate intrusions into their servers.
The company said it blocked the vulnerable path, deployed a fix and removed the injected code. It also terminated active sessions, revoked tokens, reset access for affected accounts and blocked code delivery through Cloudflare. Optimizely said it had notified all affected customers.
Helme’s checks showed the cleanup reducing the number of infected files: at 21:52 UTC on November 16, one of the 188 previously compromised scripts still contained malicious code.
Salesforce said it was investigating the modifications and, as a precaution, checking that Optimizely scripts were disabled across preview and production instances.
Wallet receipts and the broader security context
Wallets associated with the attack had received approximately $7,040, mostly in Bitcoin, by the evening of November 16, according to Report URI. Those receipts are not equivalent to confirmed losses from this incident: Helme’s investigation found that some wallets had activity predating the attack.
Hudson Rock, which tracks infostealer malware, found similar code spreading through three other web services, with an earlier wave beginning on November 13. In the Optimizely incident, Helme’s earliest observed script injection occurred early on November 16.
The firms tracing the attack had not linked it to AI tools. Optimizely attributed the intrusion to its signup-process vulnerability.
The incident followed months of concern about AI-assisted hacking after Anthropic released Claude Mythos Preview to a limited group of trusted organizations because of its ability to discover software flaws. Separately, Google’s threat intelligence team reported on November 16 that monthly vulnerability disclosures had roughly doubled between January and July.
Read also: The Binance hacker’s ETH transfers moved about $137M, but no cash-out
Article produced with the assistance of artificial intelligence and reviewed by the editorial team.
Powered by WPeMatico