What Happened
Researchers at Gambit exposed a campaign using AI agents for credit card skimming at industrial scale. A financially motivated operator chained three open-source AI tools. Strix handled scanning and flaw detection. Cairn ran autonomous exploitation. Hermes orchestrated the campaign and made tactical decisions. The human operator gave short instructions and let the agents do the rest.
Between September 10 and 15 alone, the agents launched 105 attack waves and succeeded against at least 27 companies. Overall, the campaign planted skimmers on at least 119 websites and stole more than 600,000 valid card records from two firms. Victims include a Fortune 500 hospitality company and a major US airline. Gambit estimates total AI costs between $12,000 and $18,000, an average of about $25 per target. An agent instruction to wipe card fields after theft also destroyed retailer data. Read the full report from BleepingComputer.
Why This Matters for Canadian Organizations
This campaign changes the economics of attacks on mid-sized online retailers. At $25 per target, no Canadian store is too small to be worth attacking. The attacker favoured sites running custom software, a profile common among retailers running customized Magento and headless storefronts.
The fallout lands on Canadian consumers too. Canadians buy from the same kinds of airlines, hotel chains, and online stores named as victims, and card issuers absorb the resulting fraud costs. Merchants face PCI DSS obligations and PIPEDA breach reporting when cardholder data leaves their systems. The data-wiping step adds an operational risk: a skimming incident now doubles as a data loss event.
What to Do
Monitor the integrity of checkout pages and every JavaScript file they load, including tag manager containers and CDN or S3 content. Deploy a content security policy restricting script sources. Watch for unexpected cron jobs, Kubernetes deployment changes, and database field edits in payment tables. Keep tested, offline backups of order databases. Patch custom e-commerce code and exposed admin panels on a short cycle, since AI agents find and weaponize weak points in hours.
Follow emerging attack patterns in our Trends section and daily coverage in News.
