TECHNOLOGY · AI · CYBERSECURITY · THE SYSTEMS SHAPING BUSINESS
Cybersecurity firm Gambit says an ongoing campaign targeting online retailers is using three separate AI harnesses to automate vulnerability research, exploitation and attack orchestration. The activity has been underway since July and illustrates a shift from attackers using AI for isolated tasks toward using it as connective tissue across an entire intrusion workflow.
According to the research summarized by SecurityWeek, 105 attack projects were launched between Sept. 10 and Sept. 15 alone, with at least 27 companies compromised to varying degrees. The attackers reportedly stole information associated with more than 600,000 unexpired payment cards from two victims and inserted skimmer scripts into five online stores. The campaign also reached organizations outside pure retail, including a Fortune 500 hospitality company and several U.S. businesses.
An Attack Campaign Built Around Autonomous Tooling
The details matter because the immediate event is only one part of a larger structural change now moving through enterprise technology and security.
RelatedWhy Retail Is an Attractive Proving Ground
E-commerce environments combine valuable payment data, large public attack surfaces and a web stack assembled from many third-party components. A retailer may operate a storefront platform, plugins, analytics tags, payment services, customer-support tools and marketing integrations that change frequently. That creates the kind of messy environment in which automated reconnaissance can be valuable: the attacker does not have to know in advance which software version or exposed service will produce an opening.
Once an initial foothold is established, the economics favor speed. A card skimmer can begin collecting payment details immediately, while stolen credentials can be used to explore adjacent systems. In a conventional campaign, an operator may move manually between targets. An AI-assisted workflow can keep many investigations running at once, summarize findings and route promising opportunities to the next stage. That makes relatively small criminal teams capable of behaving more like scaled operations.
Automation Compresses the Defender’s Response Window
Security teams have long relied on the idea that mass scanning is noisy and sophisticated exploitation takes time. Autonomous systems weaken that distinction. A model can read public documentation, compare an error message with known vulnerabilities, generate a test, adapt after failure and record which steps worked. Even if each individual action is imperfect, the ability to repeat those actions across hundreds of targets can produce meaningful results.
That means disclosure-to-exploitation timelines are likely to keep shrinking. Organizations that depend on a weekly patch cycle may find that a newly public weakness is already being tested before the next maintenance window. The operational response is not simply to patch faster. Teams need better asset inventories, stronger segmentation, application-layer monitoring and controls that make exploitation less valuable even when an exposed component is missed.
AI Changes the Economics of Cybercrime Services
The larger trend is the industrialization of offensive work. Criminal ecosystems already sell access brokers, phishing kits, malware, stolen credentials and laundering services. AI can reduce the expertise required to combine them. A user who does not know how to write an exploit may still be able to operate a system that identifies vulnerable targets, chooses a technique and hands off the result to another tool.
This does not mean autonomous agents have eliminated skilled attackers. High-value compromises still require judgment, persistence and operational security. But the floor is rising. Tasks that once consumed expert time can be delegated, which lets experienced operators focus on the parts of a campaign where human decision-making still matters. The result is more attempts, faster iteration and potentially broader victim sets.
The dangerous innovation is not that AI writes better phishing copy. It is that attackers can connect reconnaissance, exploitation and orchestration into a continuously running workflow.
What Defenders Should Take From the Campaign
Retailers should treat public-facing inventory as a real-time security problem. Internet-exposed systems, plugins and dependencies need owners, patch status and compensating controls that can be reviewed continuously. Payment environments deserve additional isolation because a compromise that reaches the checkout flow can be monetized quickly. Browser-side monitoring, content-security policies and integrity checks can help identify unauthorized script changes, while strong identity controls reduce the value of stolen administrative credentials.
Security teams should also expect attackers to learn from defenses more quickly. Blocking one IP address or one payload may not be enough if the orchestration layer can simply generate a new path. Detection therefore has to focus on behavior: unusual administrative actions, unexpected code changes, anomalous access to payment infrastructure and patterns that indicate automated probing across many endpoints.
The campaign is an early indicator of what cyber operations may look like when autonomous tooling becomes routine. The most important change is not cinematic artificial intelligence making independent strategic decisions. It is the mundane ability to connect dozens of small attack tasks, run them continuously and keep going when the first attempt fails. That is enough to change the economics of defense.
Retail Attacks Show What AI Changes—and What It Does Not
The campaign is important because it suggests attackers are using AI to improve throughput rather than invent an entirely new form of intrusion. Vulnerability research, target selection, payload adaptation and campaign management have always consumed time. Models can reduce that labor, allowing a small group to work more targets in parallel and to keep adapting when defenses change. The underlying weaknesses—unpatched software, exposed administration paths and weak segmentation—remain familiar, but the economics of exploiting them can become more favorable to the attacker.
That has implications for defensive prioritization. Organizations cannot respond by trying to identify whether every hostile request was generated by AI. They need controls that remain effective regardless of who or what produced the request: rapid patching, hardened administrative interfaces, strong authentication, application-layer monitoring and restrictions on data movement. AI attribution may be useful for intelligence, but resilience comes from making successful exploitation difficult and limiting the value of access when it occurs.
Retailers face an additional challenge because their systems are intentionally exposed to the public and often integrate payment, marketing, analytics and third-party applications. A compromise in one layer can quickly reach card data or customer accounts. The campaign is therefore a warning about dependency management as much as AI. Automation gives attackers more shots on goal; complex commerce stacks give them more places to aim.
What to Watch Next
The next test for AI-Enabled Attack Campaign Targets Hundreds of Online Retailers is whether defenders can turn this week's awareness into durable control changes. Security incidents create a burst of attention, but the operational value comes from what remains months later: a complete asset inventory, tighter identity boundaries, shorter patch windows and telemetry that makes abnormal behavior visible before a crisis. Those improvements are rarely dramatic enough to make headlines, yet they determine whether a similar technique succeeds the next time it appears.
Organizations should also expect attackers to reuse whatever proved economical. Once a technique works, it tends to be copied, automated and incorporated into broader criminal toolchains. Defenders therefore need to think beyond the named campaign or vulnerability and identify the underlying dependency it exposed. That may be a weak recovery process, an internet-facing management surface, an overprivileged account or a third-party connection. Fixing the specific indicator without fixing the dependency leaves the business vulnerable to the next variation.
Measurement will matter. Boards increasingly want evidence that cyber spending reduces risk rather than simply adding products. Useful indicators include time to revoke compromised access, percentage of privileged workflows protected by phishing-resistant authentication, patch latency for critical exposed systems and the share of important assets covered by reliable logging. The specific metrics vary by organization, but they force a shift from policy statements toward observable operating performance.
Finally, the response should preserve optionality. Security teams cannot know which attacker, model or exploit will dominate next quarter. Architectures that isolate systems, limit permissions and make activity auditable remain useful across many threat types. The strategic lesson from AI-Enabled Attack Campaign Targets Hundreds of Online Retailers is therefore broader than the event itself: resilience comes from designing environments in which compromise is difficult to achieve, limited in scope and fast to investigate when it occurs.
