Subvolts take AI fuels cybercrime; this episode shows how AI fights back with smarter defenses. Sep 26, 2025 - Developers
- Learn how AI models are being fine-tuned for precision in security and how to simulate attacker tactics.
- Business
- Understand the growing threat of AI-powered cybercrime and how AI can automate defenses and employee training.
This episode provides a grounded look at current AI applications in cybersecurity, discussing real-world threats and defenses.
When Deepfakes Steal $30M: The New Edge of Cybercrime
Audio streams from a16z. Your place is remembered on this device.
From Subvolts
Summary and analysis by Subvolts editorial, AI-assisted.
This AI + a16z podcast episode features Kevin Tian, cofounder and CEO of Doppel Security, discussing how AI fuels cybercrime like deepfake fraud and impersonation. Tian explains how his company uses AI to counter these threats by automating detection and takedowns, improving precision, and training employees on realistic attack tactics. The conversation explores the evolving landscape of AI-driven cybersecurity.
Cybercriminals now use AI to create fake voices and identities for scams, stealing millions. This episode explains how security companies are also using AI to detect and stop these new, sophisticated attacks before they cause harm.
This episode offers a practical look at how AI is weaponized for cybercrime and how companies like Doppel Security are fighting back. It's for security professionals and businesses concerned about AI-powered fraud. Listeners learn about the "3 V's" of social engineering and how AI models are being fine-tuned for better defense. This connects to frontier labs' work by highlighting the dual-use nature of AI, where advancements in generative models and reasoning can be applied to both attack and defense.
- Hosta16z
- GuestKevin Tiancofounder & CEO, Doppel Security
- Episode Title: When Deepfakes Steal $30M: The New Edge of Cybercrime
- Show: AI + a16z
- Length: 22 minutes
- Guest: Kevin Tian, cofounder & CEO of Doppel Security
- Attack Vectors: Impersonation, phishing, deepfake fraud
AI is transforming both sides of the cybersecurity cat-and-mouse game. Attackers are using LLMs to scale impersonation, phishing, and even deepfake fraud—while defenders are racing to automate detection and takedowns at the same speed.
In this episode, a16z partner Joel de la Garza talks with Kevin Tian, cofounder & CEO of Doppel Security (and former Uber engineer), about building in this new landscape. They cover:
- Why outsider founders sometimes build the most effective security companies
- The “3 V’s” framework for today’s social engineering attacks: volume, velocity, variety
- How Doppel uses reasoning models and reinforcement fine-tuning to cut false positives and improve precision
- Simulation tools like “vibe phishing” to train employees on real attacker tactics
- The shift from manual cyber-intelligence services to AI-driven, software-margin businesses
- Why the biggest bottleneck now isn’t model cost—but engineering time to deliver the right context
If you’re building security products or exploring how AI can automate tough edge cases, this is a ground-level look at what’s working—and what comes next.
Check out everything a16z is doing with artificial intelligence here, including articles, projects, and more podcasts.
Hosted by Simplecast, an AdsWizz company. See pcm.adswizz.com for information about our collection and use of personal data for advertising.
- LLMs
- Large Language Models
- AI systems that understand and generate human-like text.
- deepfake
- Deepfake
- AI-generated fake videos or audio that look and sound real.
- false positives
- False positives
- When a security system incorrectly flags something as a threat.
- reinforcement fine-tuning
- Reinforcement fine-tuning
- A method to train AI models by rewarding correct actions and penalizing incorrect ones.
24 minHow AI Is Rewriting the Rules of Cybersecurity | Truffle Security & Socket
WithDylan Ayrey, Feross Aboukhadijeh
47 minInside vLLM: The Engine Powering Open-Source AI
WithSimon Mo
41 minBuilding Self-Accelerating AI with Mirendil
WithBehnam Neyshabur, Harsh Mehta
43 minIdeogram’s Open-Weights Image Model and the Future of AI Design
WithMohammad Norouzi
Same guests, other shows
47 minThe good, the bad, and the future of AI agents
WithDavid Hershey
1 h 03
1 h 43The Latest in AI: Job Loss, Elon & Sam Altman Chip Race & the "AI Bubble" w/ Brian Elliott (Blitzy), Emad Mostaque & Dave Blundin | EP #197
WithBrian Elliott, Emad Mostaque, Dave Blundin
Questions people ask
- How is AI being used in cybercrime?
- Attackers use LLMs to scale impersonation, phishing, and deepfake fraud.
- How can AI help defend against these attacks?
- Defenders use AI to automate detection and takedowns, improve precision with reasoning models, and train employees with simulation tools.
Page generated Oct 19, 2026. Summaries are Subvolts' own; the audio, artwork and notes belong to a16z.