SEC536: Adversarial AI - Penetration Testing AI Systems


Experience SANS training through course previews.
Learn MoreLet us help.
Contact usBecome a member for instant access to our free resources.
Sign UpWe're here to help.
Contact Us
My work starts after something has gone wrong. A phone arrives on my desk, and my job is to find out what happened on it: who reached out to the child, when the money left, which app the conversation lived in. After more than two decades of examining criminal activity, I can tell you the hardest part of the job. It is seeing the moment in the timeline when one conversation at home would have changed everything.
October is Cybersecurity Awareness Month, and this year’s theme from the National Cybersecurity Alliance is one I would have picked myself: Don’t Make It Easy for Them. Not because families are careless. Because scammers count on getting to someone before the family has talked about what to do.
So, this month, I am asking families to have four conversations, one a week. None of them require new software. Each one takes an evening or less.
In 2025, the FBI’s Internet Crime Complaint Center received 22,364 complaints involving AI, and roughly $893 million in reported losses. Inside that category is the scam I get asked about most. A call comes in that sounds exactly like your child or grandchild, panicked and asking for money. The voice is cloned from a few seconds of audio, often pulled from a public social media video. Families reported more than $5 million in losses to these distress scams last year, and the FBI notes that the tactic is spreading beyond the classic grandparent version to impersonate a wider circle of family and friends.
In 2025, the FBI’s Internet Crime Complaint Center received 22,364 complaints involving AI, and roughly $893 million in reported losses.
The defense costs nothing. Pick one word your family agrees with in person. Anyone who calls asking for money, gift cards, or bail gets asked for the word. No word, no money. Then make the second habit automatic: Hang up and call the person back on the number you already have. These scams run on panic and speed. A callback removes both.
The National Center for Missing & Exploited Children (NCMEC) received an average of 137 reports of financial sextortion every day in 2025, up 37% from the year before. The pattern is consistent: Someone posing as a peer convinces a teenager, most often a boy, to share an image, then immediately demands money to keep it private. And generative AI has removed the one barrier that used to exist. NCMEC has documented cases where a child refuses to send an image, so the offender fabricates one and extorts them with it anyway. More than 275 identified victims in 2024 and 2025 were targeted with AI-generated abuse imagery.
The National Center for Missing & Exploited Children (NCMEC) received an average of 137 reports of financial sextortion every day in 2025, up 37% from the year before.
The single most protective thing a parent can do is say this sentence, out loud, before anything happens: "You will never be in trouble for telling me." Kids stay silent because they expect punishment, and silence is the only thing that makes this crime work. If an image is already out there, NCMEC’s free Take It Down service helps minors get explicit images removed from participating platforms. Report the account to the CyberTipline. Do not pay. Paying rarely ends the blackmail.
Common Sense Media’s national survey of 1,060 teens found that 72% have used an AI companion, and about half use one regularly. Among the teens who have used on, a third have chosen the AI over a person for a serious conversation. Nearly a quarter have shared personal information with one, including real names and locations.
I am not here to tell you AI chat is evil. I use AI every day in my work. But an AI companion is a product, built by a company, and optimized to keep your child talking. So, skip the yes-or-no interrogation and ask better questions at dinner: Which ones do you use? What do you talk about? Has one ever said something that made you uncomfortable? A third of teen users in that same survey said yes to that last one. Your kid probably has an answer. The goal is to be the person they bring it to.
I am not here to tell you AI chat is evil. I use AI every day in my work. But an AI companion is a product, built by a company, and optimized to keep your child talking.
Set aside 30 minutes, put everyone’s devices on the table. Turn on automatic updates. Put the household passwords in a password manager. Turn on multifactor authentication for email, banking, and the accounts that matter. Set the kids’ accounts to private, and while you are at it, check your own. Those public videos of your family are the raw material for week one’s voice clone.
None of this is glamorous. That is the point. The families I never met professionally did nothing fancy. They talked about it before the phone rang.
Visit Family Cyber Safety Hub for more resources to help your family stay safer online.


Heather Barnhart is the VP of Curriculum at SANS Institute, where she shapes the future of all SANS courses and trainings. She is also Dean of Faculty at the SANS Technology Institute, the premier college for cybersecurity.
Read more about Heather Barnhart