California Deepfake Law: 47 Million Views Before Deletion

Somewhere in the Philippines tonight, a parent is going to type someone's name into a search bar and pray nothing comes back. That's not dramatic — it's just where we are. The Philippine government just got Meta to promise faster removal of deepfakes and easier access to evidence in child-abuse cases, and if you've been following the fight over a california deepfake law or any of the dozen state-level bills trying to do the same thing, you already know the real question nobody's answering: faster than what, exactly?

TL;DR: The Philippines got Meta to agree to pull down deepfakes faster and share data on child-abuse cases with investigators — a real win, but "faster removal" still means the fake got out first, and that gap is where the damage happens.

What the Philippines actually got from Meta

Here's what happened, stripped of the diplomatic language. A Philippine government team met with Meta executives in Singapore, worried about how fast manipulated images and videos — deepfakes — were spreading among minors and vulnerable users. Meta agreed to two things: pull down harmful fake content quicker, and hand over data to Philippine investigators working child sexual abuse cases. According to PTV News, part of the new setup even lets individuals apply for a kind of protected status, so Meta can proactively watch for and remove deepfakes made of them, instead of waiting for a report.

That second part is genuinely new. Most platforms operate on the honor system of "tell us if you find something bad about yourself." This flips it — sort of — into "tell us you're worried, and we'll watch for you." It's not nothing. But let's not pretend it solves the core problem, which is speed versus spread. A fake doesn't wait for permission to go viral.

Why a california deepfake law offers the clearest comparison

California was one of the first states to pass a law specifically targeting AI-generated fakes, and it's still the go-to example when people argue about whether removal speed is even the right thing to measure. California law already outlaws certain non-consensual deepfake distribution, especially involving intimate images, and it is unlawful under the state's framework to spread manipulated sexual content of another person without consent. There are fines attached, and depending on the case, criminal exposure too. That's a california law restricting AI-generated content in a way that's genuinely enforceable — not just a pledge, but legislation with teeth. This article is part of a series — start with Deepfake Audio.

The Philippines deal with Meta is not legislation. It's a corporate promise, negotiated government-to-company, with no fine attached if Meta drags its feet next time. That's the difference between a state passing a bill the governor signs into enforceable law, and a company saying "we'll try to be quicker." One creates a paper trail investigators and prosecutors can use. The other creates a press release.

The math behind the california deepfake law

Malaysian authorities filed 6,987 content takedown requests between January 2024 and March 2026, and got a 95% removal rate — which sounds fantastic until you notice the second number: deepfake takedowns jumped from 817 pieces in 2024 to 3,389 in 2025, according to regional reporting picked up by Philstar.com. Removal rate went up. So did generation rate. Faster takedowns and a bigger flood aren't opposites — they're happening at the same time, on the same platforms.

That's the part that should worry you more than any single headline. It's not that platforms are slow. It's that the assembly line making fakes is outpacing the crew cleaning them up, no matter how fast the crew gets.

A single deepfake image of a public figure was seen over 47 million times before it was removed. — cited in expert research on deepfake propagation and platform response speed

Forty-seven million views. That's not a platform failing to notice — that's a platform noticing and still losing the race. A peer-reviewed analysis on PMC/NIH lays out exactly why: by the time content gets flagged, checked, and confirmed as fake, it's already been screenshotted, re-uploaded under new accounts, and mirrored on three other apps. Removal doesn't reach the copies. It never does.

Does deepfake porn count as a sexual abuse crime under state law?

In many states, yes — non-consensual sexual deepfakes are treated as a form of sexual abuse or exploitation under criminal statute, especially when the subject is a minor. Enforcement varies by state, and prosecution is still catching up to the technology, but the legal trend across the country is toward treating this as a real crime, not a prank or a gray area.

What actually protects you (and what doesn't)

If you've ever wondered whether a photo or video circulating online is really who it claims to be, that's the exact question this entire fight is about. Not "can we take it down fast," but "can anyone even tell it's fake before it spreads." Those are different problems, and right now most of the world's energy — and money — is going into the first one, because it's easier to legislate and easier to put in a press release. Previously in this series: How To Detect Deepfakes People Miss 75 Of Fake Vid.

Here's the one useful thing you can actually do tonight, before any of this policy stuff catches up: if you or your kid has a public-facing photo anywhere — school sports pages, a LinkedIn headshot, an Instagram that isn't locked down — do a reverse image search on it every few months. Not because you'll always find something. Most of the time you won't. But the earlier someone notices a fake exists, the shorter that window between upload and discovery gets, and that window is the whole ballgame. Every expert quoted in this research agrees on one thing: the real vulnerability isn't the removal speed, it's how long a fake sits unnoticed before anyone — victim or platform — even knows to look.

Where philippines deepfake removals fit the bigger picture

Zoom out and this Philippines-Meta deal is one entry in a much longer list. Australian police charged a man for making deepfakes of students, telling officers it felt "like X-ray vision," according to ABC News reporting on that case. Czech police opened 35 criminal cases tied to deepfake porn in just six months. Indian authorities booked two people for obscene deepfake videos targeting political figures ahead of an election. Every one of these is a government reacting after the fact — after the image existed, after someone saw it, after the damage started.

None of that makes the Philippines agreement worthless. Getting a platform the size of Meta to commit, on the record, to faster action and to sharing data with law enforcement on child abuse cases is a real step. It's just not the whole staircase. A state passing binding legislation — like the wave of state laws restricting AI-generated intimate images and banning undeclared bots from impersonating real people during an election or on a ballot — creates something a company pledge can't: a court can enforce it. A prosecutor can point to it. A victim's lawyer can cite it.

So here's the question worth sitting with tonight, long after you close this tab: if a fake of you started spreading right now, would you rather Meta promise to delete it in twenty minutes — or would you rather it never have been legal to make in the first place? Because right now, in most of the world, you can only pick one, and the Philippines just picked the one that arrives after the damage is already done.

Frequently Asked Questions