Six hundred thousand credit card records. Twenty-seven companies. Five days. And a total cost to the hacker of around $8,000 to $18,000, or roughly $25 per target. That is the scale of the AI credit card theft campaign that cybersecurity researchers uncovered this month, and it is quickly becoming one of the clearest signs yet that AI agents have changed what a single attacker can pull off without a team, a budget, or even much technical skill.
What Happened in the AI Credit Card Theft Campaign
Cybersecurity firm Gambit Security says it recovered the attacker’s own staging server after the operator, believed to be a Chinese–speaking individual, left it exposed on the open web. That mistake let researchers reconstruct the entire campaign step by step. Between September 10 and 15, 2026, the operator launched 105 attack projects and compromised at least 27 companies, including a Fortune 500 hospitality brand, a major U.S. airline, a large industrial supplies distributor, and an online fashion retailer. The activity reportedly began back in July and, according to researchers, is still ongoing.
According to Gambit, the operation stole upward of 600,000 valid credit card records from two of the compromised companies, with the large majority belonging to U.S. cardholders, and skimmer code was found on at least 19 confirmed victim sites and more than 100 additional websites. Cloudflare has confirmed it has repeatedly taken down servers tied to the campaign, though the attacker keeps rebuilding them.
How Three AI Tools Did the Work of a Hacking Team
What makes this campaign different from a typical data breach is who, or rather what, did the actual hacking. The operator relied on three off-the-shelf, open-source AI agent frameworks, each handling one stage of the attack:
- Strix scanned potential targets and identified security vulnerabilities worth exploiting.
- Cairn handled the actual break-in, working autonomously to gain access once a weakness was found.
- Hermes orchestrated the wider campaign and ran post-access tasks, including planting card skimmers and, in some cases, wiping evidence afterward.
According to researchers, the human operator typed fewer than 2,000 short prompts across 260 sessions, often just a handful per target, while the agents carried out the scanning, exploitation, and cleanup almost entirely on their own. In at least two cases, the AI cleanup routine reportedly went further than intended and wiped out victims’ own backup data along with the evidence it was told to erase.
Why This Attack Only Cost About $25 Per Target
Traditionally, breaching dozens of companies at this scale would have required a team of skilled hackers working for weeks. Gambit Security says the AI models were rented through OpenRouter, a marketplace that provides access to various AI systems, and that the entire campaign cost an estimated $12,000 to $18,000 in AI usage. Spread across the targets, the average completed scan cost just $25.46, ranging from about $3 for the cheapest target to just under $80 for the most expensive.
That price tag is the real headline here. A capability that once demanded a skilled team and significant time now costs less than a nice dinner per target, and can run at machine speed against dozens of companies a day.
5 Signs Your Card Details May Have Been Skimmed
Card skimmers work quietly in the background of a checkout page, so there is rarely an obvious sign at the moment of purchase. Instead, watch your accounts for these patterns:
- Small test charges. A charge of a dollar or two from an unfamiliar merchant, often the first thing criminals do to check whether a stolen card still works.
- Purchases you never made, especially from retailers or in countries you have no connection to.
- A new card suddenly declined for no clear reason, which can happen after a bank flags unusual activity on its own.
- Unexpected password reset or login alerts from a store where you have an account, which may point to broader access beyond just your card.
- A notification from the retailer itself about a security incident, weeks or months after you actually shopped there.
What to Do If You Think Your Card Was Exposed
If you shopped at a large retailer, airline, or hotel brand recently and want to be safe rather than sorry, a few simple steps go a long way:
- Review your last two or three statements line by line rather than skimming the total.
- Call your bank and request a new card number if anything looks off, rather than waiting to see if more charges appear.
- Turn on purchase alerts through your banking app so new charges reach you the moment they happen.
- Use a virtual card number for online shopping where your bank offers one, so a compromised number cannot be reused elsewhere.
- Avoid saving your full card number on retailer websites when you can, especially smaller or less familiar stores.
What This Means for Online Shoppers Going Forward
This campaign is part of a wider pattern this year of AI agents being turned toward real-world crime rather than just research demos. Our earlier coverage of AI voice cloning scams looked at a similar shift: a threat that used to require real skill and real effort has become cheap and automated enough that far more criminals can attempt it. The lesson for shoppers is not to stop buying online, but to treat routine account monitoring the same way you treat locking your front door: not exciting, but the thing that actually keeps you safe.
For businesses, security researchers point to unglamorous but effective fixes: patching known web application vulnerabilities, auditing who has admin level access, scanning for exposed cloud credentials, and monitoring checkout pages for unauthorized code changes in real time, since that is exactly where these skimmers get planted.
The Bottom Line on AI credit card theft
The AI credit card theft campaign is a preview of where cybercrime is headed: fewer skilled attackers, cheaper attacks, and more of them running at once. According to Cybernews, remediation for breaches like this often takes companies weeks, while the attackers themselves are finished within hours. Until security catches up to that speed gap, the most reliable protection is still the boring kind: check your statements, turn on alerts, and treat every online checkout page as one your bank, not just the retailer, needs to be watching too.
Frequently Asked Questions
How many credit cards were stolen in the AI credit card theft campaign?
Researchers at Gambit Security say the campaign stole more than 600,000 valid credit card records, with roughly 488,000 belonging to U.S. cardholders.
How much did the attack cost to carry out?
The operator spent an estimated $12,000 to $18,000 total on AI model access, averaging about $25 per completed attack on a target company.
Is this AI credit card theft campaign still active?
As of late September 2026, researchers say the campaign is still running, and the attacker has repeatedly rebuilt infrastructure that Cloudflare has taken down.