Ethan LockshinSales at NVIDIA focused on AI adoption

AI hasn’t changed scam methods—just boosted speed and scale, making fraud easier and more efficient

Name: AI hasn’t changed scam methods—just boosted speed and scale, making fraud easier and more efficient
Uploaded: 2025-12-08T15:27:00+00:00
Duration: 1 min 37 s
Channel: Ethan Lockshin
Description: AI isn't creating new scams, just making old ones faster and cheaper. Phishing campaigns are now hyper-personalized at scale, and ransomware variants run…

8 Dec 2025 · 2 min

Watch on YouTube

1:37

AI isn't creating new scams, just making old ones faster and cheaper. Phishing campaigns are now hyper-personalized at scale, and ransomware variants run AI models on infected devices. The real danger: attackers exploiting trust in automated security tools through prompt injection and context poisoning.

Chapters

What new frontiers like AI phishing or LLM poisoning or supply chain AI attacks will shape the next phase of security operations and how mate is preparing for them.
So I don't think that AI driven phishing, it's the new frontier. What I mean by that, I'll explain is that AI driven phishing is a faster way of doing phishing, a cheaper kind of very spare headed phishing and very personalized. But the fundamental defense and the fundamental way of attacking has not changed much for the most part. Of course there's like new variants of ransomware that are now running local LLMs on devices, which is great, but still 90 or even more percent of what's happening outside is a lot of using the same techniques, just doing them faster. So the economics have changed. Attackers can now run sophisticated campaigns at a massive scale with minimal effort. They also, at least not the nation state, but for, for sure for the financial groups they operate for roi. Right? And the threat vectors that I'm watching closely are the ones that exploit trust in automation. So it can be context poisoning. So I can already know that you might be using an invariant of AI soc. And I can prompt injections and other type of variants, even invisible ones, to my emails or to PowerShell scripts or to command lines that I know the AI SOC will utilize, right? Or whether it's an AI SoC or an analyst copying stuff and putting them in ChatGPT or Gemini or Claude. And I think it's a lot about really utilizing similar techniques as we've seen in the past, at scale for a cheaper price.