Loti.

Somewhere right now, a server farm is looking at your face. Not a person - a machine, churning through a day's worth of uploads to the public internet, asking one quiet question over and over: is this you, and did you say it could be here? That machine belongs to Loti. And the people it works for range from Oscar nominees to someone whose ex posted something they never agreed to.

Loti AI is a Seattle company that finds unauthorized copies of your face and voice online and gets them taken down. Deepfakes. AI clones. Impersonation accounts. Leaked photos. The scammy ad with a celebrity who never endorsed anything. Loti scans for it, confirms it is unauthorized, and files the takedown - mostly without a human lifting a finger. The promise is almost old-fashioned in its simplicity: your identity belongs to you.

The Problem They Saw

When a face became copy-paste

For most of history, the worst thing someone could do with your likeness was draw it badly. Generative AI changed the math. Today a convincing fake of you - moving, talking, selling crypto - can be made in minutes by anyone with a laptop and a grudge. The supply of fake faces went vertical. The tools to defend a real one stayed roughly where they were: a frantic email to a platform, a takedown form, a wait, a shrug.

The founders noticed the asymmetry early. Fakes scale; defenses don't. A single person, however motivated, cannot manually patrol an internet that uploads hundreds of millions of items a day. You cannot win a speed contest against software by typing faster. The only honest answer to automation is more automation - pointed the other way.

The Founders' Bet

A side project that refused to stay small

Loti began in 2022 as a favor. Luke Arrigoni - a Columbia math graduate who had run an AI consultancy working with Getty Images, Johnson & Johnson's Janssen, UPS and AT&T - and his wife Rebekah started it to help one person scrub unwanted images off the web. It was the kind of weekend problem that seems narrow until you look closely and realize it has no floor.

Then 2023 happened. The Hollywood strikes put AI likeness rights on every marquee in town, and Luke's industry contacts came calling with the same worry, urgently. The side project acquired a co-founder and CTO in Hirak Chhatbar, a thesis, and a company name. The bet was specific: that protecting a likeness is not a celebrity luxury but a universal need, and that the same engine could serve a global pop star and a schoolteacher without redesigning either.

The Product

Forty-odd models and a very patient bird

Under the hood, Loti runs roughly 40 to 50 machine learning models built for synthetic-media detection and face-and-voice matching. It needs only a handful of reference photos to learn you, and it can still recognize you when half your face is hidden. The system runs on a network of tens of thousands of servers that, together, work through a full day's worth of internet uploads inside 24 hours. When it finds something unauthorized, the takedown engine moves on it - leaning on copyright, the right of publicity, and Tennessee's ELVIS Act, a statute named, fittingly, after Elvis Presley.

The lineup is deliberately tiered. There is a free plan for everyone, with 24/7 monitoring of your face and voice. There is Watchtower, the paid tier for public figures who need scale and speed. And there is an IP-protection API, aimed at managing likeness and brand rights centrally - useful when the thing being faked is worth licensing in the first place.

The Proof

Numbers that do the arguing

Skepticism is the correct posture toward any company that promises to police the internet. So look at what is measurable. Loti reports clearing flagged content with roughly a 95% success rate, usually within about 17 hours - the gap between "I found a fake of myself" and "it's gone" measured in a single workday rather than a legal quarter.

The other proof is who shows up. Talent agencies WME and CAA route their rosters to Loti. Entertainment firm Dolphin extended access across subsidiaries including 42West, The Door, Shore Fire and Elle Communications. In January 2026, Underscore Talent signed on to protect global creators. And the money agrees: a $16.2M Series A led by Khosla Ventures, on top of seed backing from FUSE, Bling Capital and Ensemble, for about $23M total.

The Mission

Putting people at the center of AI

Loti frames its work in terms most AI companies avoid: restraint. The stated mission is to give individuals effortless control over their digital identities - to put people, not models, at the center. The number it likes to cite as a north star is zero: zero images of you online that you never approved. It is an asymptote, not a finish line, and Loti seems to know it. The point is the direction of travel.

There is an irony worth naming. The same technology that makes the fakes - face recognition, voice modeling, generative matching - is the technology Loti turns around to catch them. The arsonist and the fire brigade shop at the same store. Loti's wager is that the brigade can be faster, and that being faster is most of the battle.

Why It Matters Tomorrow

The next normal

As generative tools get cheaper and better, the question stops being whether a fake of you exists and becomes how quickly you can make it stop. That shifts likeness protection from a celebrity service into ordinary infrastructure - something closer to spam filtering than to a bodyguard. Loti is betting it can be the layer that does it quietly, in the background, for anyone.

So return to that server farm, still looking at your face. The difference Loti is trying to make is not that the looking stops - the internet's looking never stops. It's that, for the first time, something is looking back on your behalf, all day, and pulling down the fakes before you ever have to see them. The machine doesn't sleep. That, for once, is the comforting part.