The Zero-Marginal-Cost Fraud Economy
Generative AI did not just build faster code assistants. It broke the economics of fraud.
Back when I was writing term sheets for enterprise cybersecurity plays, corporate defenses were all about perimeter security. We built high walls for Fortune 500 companies because that was where the value concentrated. Attackers spent months researching targets and scripting custom exploits. The math worked. High effort was justified by a high payout. It did not make economic sense to target individual consumers with bespoke social engineering campaigns; the ROI was non-existent.
Cheap large language models and instant voice-cloning APIs changed all that. Now, a threat actor can clone your daughter's voice from a three-second snippet pulled off a public Facebook video. They can automate caller ID spoofing and instantly look up local geography. High-end social engineering is no longer craft work. It is factory work. Attackers have scaled it into a volume game targeting everyday consumers. Imposter scams stole $3.5 billion in 2025 alone, triple the amount from 2020, according to FTC data. This is not just an older generation vulnerability, either. Gen Z gets targeted constantly by SMS scams, and research from Malwarebytes indicates they fall for them a quarter of the time. The perimeter has moved from the enterprise network to the family group chat.
Coughlin's Pedigree and the Seven Million Dollar Syndicate
This shift in threat dynamics is the operational background for Savi Security's new launch. The startup just popped out of stealth with a $7 million seed round led by Acrew Capital, with participation from Resolute Ventures, TTCER, and Magnify Ventures.
But the real draw is the founder pedigree. Patrick and Ryan Coughlin are seasoned product and security operators. Patrick spent years in federal cyber defense before co-founding TruSTAR, which sold to Splunk for $82 million in 2021 (before Cisco snapped up Splunk in 2024). He ended up serving as Cisco's VP of security products. Ryan built consumer-facing products at Apple and Spotify.
The genesis of the company was highly personal. Two years ago, their mother received a call. The caller ID showed her daughter's number. When she answered, she heard a voice scream "Mom, they've got me." A man came on, demanding $1,200, referencing the parking lot of the local Walmart she frequented. The voice was a cloned copy; the details were scraped. She kept her cool and called her daughter directly to verify she was safe, but the incident shook Patrick. He realized the same nation-state level of sophistication he was defending against at Cisco was now targeting his mother for twelve hundred bucks.
The Family Plan Unit Economics Playbook
Consumer security is a hard venture sector. The customer acquisition costs (CAC) are notoriously high because people rarely buy security tools until they have already been breached. When they do buy, they churn fast.
Savi is attacking this go-to-market blocker with a pricing structure designed for viral, family-wide distribution rather than individual sales. They charge $8 a month, or $63 if billed annually. The clever part is that the subscription covers an entire family with no limit on the number of users.
For a VC looking at this, it is a smart play to solve the CAC problem. By targeting the "designated system administrator" of the family—the tech-savvy sibling or child—Savi gains an advocate who will install the app on five or six family devices. This drives down the dynamic CAC per seat and increases retention. People will churn off a personal utility, but they will hesitate to cancel a tool that protects their aging parents or children. The key variable is whether the backend API costs of running continuous AI text and call scanning will eat their gross margins.
Live-Call Monitoring as a Tech Moat
What is the actual product that justifies the annual subscription? Basic features screen incoming text messages and voicemails for phishing scams. That is table stakes; standard antivirus suites and mobile OS built-ins already screen basic SMS spam.
Savi's primary competitive moat is its live-call monitoring capability. If a user receives a suspicious call, they can opt to add Savi's active listener to the call. The app monitors the conversation in real-time, scanning for behavioral indicators of fraud—like high-pressure ransom demands, artificial speech cadences, and sudden cash-out instructions.
Building a model for this requires massive amounts of real-world training data. Four months ago, the founders launched Scamwise, a free, anonymous website where users could upload fishy emails and call recordings. The site gathered 100k submissions, according to Savi's TechCrunch interview. That proprietary dataset is what trains Savi's detection model. The app runs on an AI gateway—leveraging Google's Gemini as the primary engine but maintaining the ability to route traffic to voice-specific models. It is a personal security operations center (SOC) for people who do not have an IT department to call.
Bypassing the Social Engineering Trust Loop
Savi's launch highlights a broader transition in cybersecurity. Cybercriminals have realized that they do not need to exploit software bugs if they can exploit human relationships.
We have seen this trend in the enterprise space. Attackers build fake reputation networks across GitHub and YouTube to distribute malware. As we analyzed in our coverage of Reputation Hijacking in Crypto Attacks, threat actors are exploiting trust signals that people naturally depend on. They generate fake positive reviews and manipulate platform metrics to make malicious links appear safe.
The same psychology is at play in consumer voice scams. When your phone displays your daughter's caller ID, and a voice speaking in her exact tone screams for help, your critical thinking shuts down. The trust is pre-established. The risk is massive, and resolving this challenge requires new approaches. While corporate security leaders grapple with enterprise threats, as detailed in AI’s Dual Threat: Complexity and the CISO Capability Gap, the consumer front line remains even more vulnerable because they do not have a dedicated security team checking their calls. Savi's approach bypasses the trust loop entirely by monitoring corporate-level fraud patterns within consumer-level calls.
Is Consumer Security a Venture-Scalable Asset Class?
Can Savi scale this into a venture-backed powerhouse, or will it remain a niche utility? The addressable market is large, but compute economics will decide the outcome.
Running real-time voice inference and behavioral models across millions of calling minutes gets expensive fast. By building their software on an AI gateway, Savi can optimize routing to the cheapest model that gets the job done. That protects their gross margins. But they will need to convert their early momentum from the Scamwise project into paid subscriptions quickly to offset the compute bills.
If they can successfully scale the family plan model, the dataset network effects will protect them. Every scam voice model they identify in the wild makes their detection stronger. In the generative AI era, security isn't about building bigger firewalls. It is about building faster data loops. Savi has the pedigree and the capital to prove if consumers will pay for that loop.