Why Hack When You Can Ask?
Stay updated with us
Sign up for our newsletter
An extortion group called BlackFile has been in the security news lately, targeting retail and hospitality organizations with a campaign that began in February 2026. Researchers at the Retail & Hospitality ISAC have detailed how the group steals data and demands seven-figure ransoms, but what’s most notable is how BlackFile gets in.
Rather than relying on custom malware or sophisticated pieces of code hidden inside phishing emails, BlackFile starts with a phone call. Attackers impersonate a corporate IT help desk, dial an employee, create a sense of urgency, and walk that employee through handing over credentials and one-time passcodes. From there, the attackers register their own devices, bypass multi-factor authentication, scrape internal directories, escalate to executive accounts, and exfiltrate data through the same APIs and channels that the organization uses every day. The technical footprint is nearly invisible because the attack isn’t technical.
Read More: When Telecom Infrastructure Starts to Behave Like Enterprise Software
This Isn’t New, But It’s Getting Worse
BlackFile isn’t an anomaly. High-profile attacks from groups like Scattered Spider and Shiny Hunters in recent years have provided a dramatic illustration of how successful this playbook can be. They combined basic reconnaissance, impersonation, and social engineering to breach some of the most well-defended companies in the world. What we’re witnessing now is how that playbook has been copied, refined, and industrialized by a growing number of threat actors who have recognized that people are easier to manipulate than systems are to compromise.
The retail and hospitality sectors are particularly exposed. These industries run on large, distributed workforces, high call volumes to IT and service desks, and frontline staff who are trained to be helpful and responsive — qualities that social engineers deliberately exploit. When an employee gets a call from someone who sounds like IT, knows internal terminology, and creates a sense of urgency around an account lockout or a compliance deadline, the instinct to help is not a failure of judgment. It’s exactly how these employees are supposed to behave.
AI Is Making This Even Worse Before It Makes It Better
The threat is accelerating. Generative AI has dramatically lowered the barrier to executing believable social engineering attacks at scale. Attackers no longer need to be fluent in English to impersonate a help desk technician convincingly. Voice cloning tools can replicate an employee’s speech patterns from a few seconds of publicly available audio. Large language models can research a target organization, construct a credible pretext, and generate natural conversation scripts faster than any human red team could.
What we’re seeing from groups like BlackFile and the broader network researchers have linked them to — a loose-knit collective sometimes called “The Com” — looks less like traditional hacking and more like industrialized deception. Farms of human attackers, augmented by AI tools, are working through target lists and conducting live conversations designed to extract credentials, approvals, and access. The scale is new. The attack surface — human trust — is not.
Reactive Training Isn’t a Defense Strategy
When social engineering attacks succeed, the instinct is to blame the employee and schedule more training. It’s understandable. Security training as a response is available, auditable, and actionable. But training as a strategy has a fundamental flaw: it asks employees to become suspicious of the urgency, authority, and helpfulness that their jobs require them to extend every day.
Read More: The Agentic SOC: Why Security Operations Must Reimagine Itself—and Fast
When an agent gets on with a live caller who knows the employee’s manager’s name, references a real internal ticket number, and creates time pressure around a plausible scenario, it’s unreasonable to expect the agent to do anything other than attempt to help. It’s human nature. Urgency disrupts critical thinking. Authority reduces resistance. And the desire to avoid being the person responsible for harm is a powerful motivator to do what’s been asked. No amount of annual training can reliably overcome those dynamics in the moment.
A New Path: Detection and Response, Applied to People
In virtually all cybersecurity disciplines, we follow a common pattern to protect ourselves: detection and response. From endpoint to network to cloud, we apply continuous monitoring technologies to the assets we’re trying to protect. However, for most of cybersecurity’s history, social engineering was treated differently. Human interactions are extremely complex and highly dimensional, so the training was the best we could do (even though we knew it was insufficient).
That’s changing. Advances in conversational AI and natural language processing are making it possible to analyze live interactions — across voice, video, chat, and service desk channels — for the behavioral signals that characterize social engineering: manufactured urgency, authority exploitation, procedure evasion, and impersonation.
The same AI technology that attackers are using to scale their campaigns can now be applied defensively. It can detect when an interaction bears the hallmarks of an attack and give security teams the ability to intervene before credentials change hands or access is granted.
Protect People, Don’t Punish Them
BlackFile-style human exploits will continue to grow. The economics of human-layer attacks are too favorable and the historic defensive gap too wide. At its core, every organization is a group of people, and those people are intrinsically vulnerable. No amount of training can make that go away.
However, there is hope. Enforcing MFA for identity verification wherever possible helps. And by pairing that best practice with the capabilities of language models, we can treat these attacks as a detection and response problem — the same framework we’ve applied to every other category of threat — rather than a training problem we can solve with one more awareness module.
As an industry, we’ve spent years trying to train away the very trust and helpfulness that makes us all human. It’s time to stop blaming the victim and start protecting our people.
