Savi Security launched a consumer scam-defense app on iPhone and Android on Tuesday, backed by a $7 million seed round led by Acrew Capital with participation from Magnify Ventures, TTCER, and Resolute Ventures. The startup, founded by brothers Patrick and Ryan Coughlin, screens texts, voicemails, and live phone calls for AI-generated fraud — including voice clones that impersonate family members. Its headline feature is a live-call listener that a user can add during a suspicious conversation to flag behavioral tells in real time.
The company was built out of a specific incident. About two years ago, Patrick Coughlin's mother received a call showing her daughter's caller ID, followed by what sounded like her daughter's voice pleading, a scream, and a man demanding $1,200 or he would kill her outside the local Walmart she frequented. She called her daughter, who was unharmed. The kidnapping was fabricated end-to-end with AI — the voice, the number spoofing, and the location detail.
Coughlin was running security products at Cisco at the time, having landed there after Splunk acquired his cloud security startup TruSTAR for a reported $82 million in May 2021, and after Cisco bought Splunk in 2024. His question after the call was why the sophistication he had spent a career defending governments and Fortune 500s against was now being aimed at his mother.
Key facts
- 01Savi Security raised $7M in seed funding led by Acrew Capital, with Magnify Ventures, TTCER, and Resolute Ventures participating.
- 02The app costs $8/month or $63/year and covers unlimited family members on a single plan.
- 03The FTC says imposter scams cost consumers $3.5B in 2025, triple the 2020 figure.
- 04Savi's free Scamwise site logged 50,000 submissions in four months and now takes 10,000+ per week, hitting 100,000 total on Tuesday.
- 05A voice can be cloned from roughly three seconds of publicly available audio.
The economics of consumer fraud have inverted. Before generative AI, running a targeted scam on an individual required custom research, voice tooling, and time — the math only worked against enterprises and governments. Cheap large language models, off-the-shelf voice cloning, and scrapable social media footprints have collapsed the cost to near zero, and the addressable target list has widened to everyone with a phone.
Coughlin's point on how thin the audio requirement has become is worth reading in full.
The FTC reported last month that Americans lost $3.5 billion to imposter scams in 2025, triple the 2020 figure. Older consumers file the majority of reports, but a 2025 study from Malwarebytes found Gen Z is targeted with text scams more often than other cohorts and falls for them roughly 25% of the time. The demographic assumption that younger, more online users are harder to trick has not held up against generative-AI-quality lures.
Savi test-drove its detection model with a free website called Scamwise, where anyone can upload a suspicious text, photo, or email anonymously and get a verdict. Launched about four months ago, Scamwise pulled in 50,000 submissions in its first stretch and now receives more than 10,000 per week, hitting 100,000 total on Tuesday. That flow doubles as in-the-wild training data for Savi's model.
Under the hood, Savi runs primarily on Google Gemini but is wired through an AI gateway that lets it swap in specialized models — voice-detection systems in particular — as they become available. That architecture matters for a defense product: the offensive side rotates through models constantly, and a defender locked to a single provider ages out fast.
Pricing is unusual. Savi charges $8 per month, or $63 per year, and the plan covers an entire family with no cap on users — kids, spouse, parents, and any relative the account holder wants to add under administrative oversight. That structure implicitly acknowledges that the weakest link in a family's scam defense is rarely the person paying the subscription.
The category is crowded — Malwarebytes and others offer text and call screening — and Savi's live-call monitoring will be judged on false-positive rates and latency, both of which are unforgiving in a real phone conversation. Consumer security products also face a long-running adoption problem: people install them after they get hit, not before. The Scamwise funnel is Savi's attempt to solve that, converting the free-tool audience once they've seen a verdict on something aimed at them.
Coughlin's broader read is that AI is not just arming existing criminal syndicates but pulling amateurs into fraud by lowering the skill floor. That expansion of the attacker pool is the harder problem for defenders — syndicates are finite and trackable, ad-hoc opportunists are not. A real-time consumer layer that runs on-device during the call itself is a plausible response to that shape of threat, in the way endpoint antivirus was to an earlier one.
The bet for Savi is whether consumers will pay a monthly fee for something that, when working, produces no visible output. That has historically been a hard sell outside of enterprise security. What has changed is that the threat now arrives in a parent's voice on a spoofed number — and the $1,200 ransom demand at a specific Walmart is no longer a hypothetical pitch deck slide.
Working on something we should cover, or seeing a story we missed? Send leads, documents, or feedback to hello@aichatdaily.com. For sensitive tips, see our secure tips page for Signal and PGP options.
Spotted an error? Email hello@aichatdaily.com with the URL and the issue, or read our full corrections policy.
