For years, the internet operated according to a brutal rule:
Once an image was online, getting it back was almost impossible.
Then artificial intelligence made the problem worse.
Now the image doesn’t even have to be real.
A face can be taken from a school photograph, an Instagram account, a professional profile, or an ordinary family picture and placed onto sexually explicit material that never existed.
No encounter.
No camera.
No consent.
Yet to everyone who sees the finished product, it can look horrifyingly convincing.
That is the nightmare lawmakers were trying to confront with the TAKE IT DOWN Act.
Its emergence represented something unusual in Washington: politicians from opposing parties agreeing that technological change had created a form of abuse serious enough to demand a federal response.
The central principle was straightforward.
Creating or distributing certain intimate images without consent—including qualifying AI-generated or digitally manipulated explicit images—should carry consequences.
And when victims discover such material online, covered platforms should not be allowed to shrug indefinitely while it spreads.
That matters because intimate-image abuse operates at internet speed.
Seventy-two hours can sound like an arbitrary deadline until you imagine being the person inside the photograph.
You wake up to messages.
Someone from school sends a link.
Then another person.
Then a coworker.
Your name begins appearing in searches.
Strangers save copies.
Accounts repost them.
Someone creates a compilation.
Someone else adds your address or employer.
You report the original post.
Nothing happens.
You report it again.
An automated system answers.
The image continues moving.
By lunchtime, thousands of people may have seen something that depicts you naked even though you never posed for it.
By evening, copies may exist across multiple services.
At that point, the distinction between “real” and “fake” offers surprisingly little emotional protection.
Your body may not have been photographed.
Your dignity was still attacked.
That is what older legal frameworks have sometimes struggled to capture.
Deepfake sexual abuse creates a strange contradiction.
The depicted event never happened.
The harm absolutely did.
A teenager can know intellectually that the image is fabricated and still dread walking into school.
A teacher can understand that a video is synthetic and still fear parents or administrators seeing it.
An employee can explain the manipulation and still wonder what coworkers now believe.
A victim can repeat, “That isn’t me,” a hundred times and still feel violated because the face is theirs.
The humiliation attaches to a real person.
That is why calling these creations “fake porn” can sometimes minimize what victims experience.
The sexual act may be fictional.
The targeting isn’t.
Someone selected a person.
Obtained their image.
Manipulated it.
Distributed it.
And often relied on the internet’s most predictable instinct:
People will look before they verify.
For victims, removal therefore matters enormously.
Punishment months or years later may provide accountability, but it does not stop tonight’s repost.
A meaningful response has to address both sides.
Who created or distributed the material?
And how quickly can its circulation be interrupted?
The TAKE IT DOWN framework attempted to confront that second problem by requiring covered platforms to establish a process through which victims could request removal of qualifying nonconsensual intimate imagery.
The 72-hour requirement became one of its defining features.
The philosophy behind it is important.
Platforms cannot necessarily prevent every abusive upload.
But once properly notified, they should not be indifferent to it.
That represents a significant shift in how society thinks about online responsibility.
For decades, technology companies often described themselves primarily as intermediaries.
Users create the content.
Users post it.
The platform provides infrastructure.
But infrastructure has power.
A recommendation system can turn one malicious upload into thousands of views.
Search can make humiliation discoverable.
Sharing tools can multiply it.
Notifications can push it toward people who never requested it.
The same systems that make harmless content travel efficiently can make abuse travel efficiently too.
So the difficult question becomes:
When does hosting become responsibility?
There is no simple answer.
Make platforms responsible for everything users post, and enormous portions of the open internet become difficult to operate.
Give platforms no responsibility at all, and victims can become trapped inside systems where nobody has sufficient incentive to act quickly.
The challenge is designing obligations narrow enough to target genuine abuse without creating a mechanism that can easily be weaponized against lawful expression.
That is where the free-speech concerns enter.
And they should not simply be dismissed.
A removal system powerful enough to erase unlawful material can potentially be misused to demand removal of lawful material.
People can submit false claims.
Satire can be misunderstood.
Newsworthy material can involve intimate subjects.
Evidence of misconduct can raise difficult questions.
Automated moderation can make mistakes.
Platforms facing legal risk may remove first and investigate later.
That phenomenon is sometimes called over-removal or over-censorship.
From a company’s perspective, deleting questionable content may feel safer than risking liability.
From society’s perspective, however, a system that routinely removes lawful speech because companies are frightened of penalties creates another problem.
This is why definitions matter enormously.
What counts as an intimate image?
What constitutes consent?
What qualifies as a digital forgery?
What evidence must accompany a removal request?
What happens when the uploader disputes the claim?
What safeguards exist against fraudulent notices?
What happens to material with legitimate public-interest value?
The quality of a law is often determined in these unglamorous details.
“Protect victims” is an easy principle to support.
Building a system that protects victims accurately, quickly, and consistently is much harder.
Still, free expression has never meant an unlimited right to sexually exploit another person’s identity.
Society already recognizes categories of conduct where speech and imagery intersect with privacy, harassment, threats, fraud, defamation, exploitation, or other legally cognizable harms.
AI does not erase those concerns.
It intensifies them.
The old assumption was that convincing photographic evidence required an underlying event.
Someone had to stand in front of a camera.
Someone had to be there.
Generative technology breaks that relationship.
A believable image can now be manufactured without the depicted person’s participation.
That changes the economics of humiliation.
Once, creating a sophisticated fake required specialized skills.
Increasingly, the barrier is much lower.
A person with ordinary photographs and accessible software can potentially manufacture explicit material targeting someone who rejected them, embarrassed them, defeated them in an election, competed with them professionally, or simply happened to attract their attention.
Revenge becomes scalable.
So does bullying.
Imagine what that means inside a high school.
A teenager takes a classmate’s photograph.
Minutes or hours later, explicit fabricated images begin circulating among students.
The victim may be fifteen.
The creator may also be fifteen.
The technology is new.
The cruelty is ancient.
But the scale is different.
A cruel rumor once moved through hallways.
A synthetic image can follow someone home.
It can appear on phones at midnight.
It can be saved indefinitely.
It can resurface years later.
The victim’s future college, partner, or employer may encounter it without understanding its origin.
That permanence changes the psychological stakes.
Adults are vulnerable too.
An abusive former partner can use synthetic imagery as another mechanism of control.
A stalker can fabricate sexual material.
A woman running for office can be targeted before an election.
A journalist can be humiliated after publishing an investigation.
A business owner can be threatened with fabricated images unless money is paid.
A public figure’s face can be inserted into explicit content at enormous scale.
The technology doesn’t care why someone is being targeted.
It simply reduces the effort required.
This is why bipartisan agreement became possible.
Deepfake intimate abuse does not fit comfortably into traditional ideological categories.
A conservative daughter can be targeted.
A progressive activist can be targeted.
A Republican politician.
A Democratic politician.
A teenager who has never thought about politics.
A teacher.
A soldier.
A nurse.
Anyone whose photograph exists can potentially become raw material.
That universality can clarify the moral issue.
Consent does not become obsolete because pixels were generated instead of photographed.
If anything, synthetic media forces society to define consent more carefully.
A person may consent to posting a selfie.
That is not consent to having the selfie transformed into pornography.
A person may consent to sending an intimate image to one partner.
That is not consent to public distribution.
A person may have once consented to an intimate photograph.
That does not necessarily mean they consented to every future use of it.
Digital technology encourages copying.
Human dignity still requires boundaries.
The law’s strongest supporters see those boundaries as overdue.
For victims, the ability to point to federal protections can carry psychological importance beyond the mechanics of litigation.
It says:
This happened to you.
The fact that the image was synthetic does not make your violation imaginary.
You are not overreacting.
You do not have to accept this as the price of having a face online.
That recognition matters.
Victims of technology-enabled abuse are often told to solve the problem themselves.
Delete your account.
Stop posting photographs.
Make everything private.
Ignore the trolls.
Don’t read the comments.
Change your name.
Change schools.
Change jobs.
In other words, the victim is asked to disappear while the perpetrator keeps participating.
That is backwards.
The person targeted should not automatically carry the burden of retreat.
But legislation alone will not solve the problem.
Even the strongest removal requirement encounters a basic feature of the internet:
Copies multiply.
An image removed from one service can remain on another.
Someone can download it before deletion.
Private groups can redistribute it.
Foreign websites may operate outside straightforward enforcement.
Encrypted channels create additional complications.
A successful takedown can reduce exposure dramatically without guaranteeing total disappearance.
Victims deserve honesty about that.
“Removed” does not always mean “erased from existence.”
That is why prevention, education, platform design, enforcement, and technological detection all matter alongside law.
AI developers have responsibilities too.
Systems capable of generating realistic human imagery should be designed with abuse risks in mind.
Platforms should invest in reporting mechanisms that traumatized users can actually navigate.
Law enforcement needs sufficient understanding of synthetic media to distinguish fabrication from authentic evidence.
Schools need policies for AI-generated sexual harassment.
Parents need to understand that teenagers do not require advanced technical expertise to encounter or misuse these tools.
And society needs to stop treating digitally mediated abuse as somehow less real because it happened through a screen.
The screen is simply the delivery mechanism.
The consequences occur in human nervous systems.
Human classrooms.
Human workplaces.
Human marriages.
Human reputations.
Human lives.
At the same time, defenders of civil liberties are right to keep examining how takedown powers operate.
Good intentions do not immunize laws from unintended consequences.
History contains plenty of examples of tools introduced for narrow purposes and later applied more broadly.
That is why oversight should continue even after legislation passes.
Are legitimate requests being processed quickly?
Are false requests being rejected?
Are platforms removing lawful material unnecessarily?
Are victims able to appeal failures?
Are accused uploaders afforded appropriate protections?
Are smaller platforms capable of complying?
Are bad actors learning to exploit loopholes?
Legislation should be evaluated by what happens after the signing ceremony.
The real test is the terrified person staring at a screen at two in the morning.
They have just discovered an explicit image bearing their face.
Maybe they’re sixteen.
Maybe they’re sixty.
They know it isn’t real.
They also know hundreds of other people may not.
What happens next?
Can they find the reporting mechanism?
Does someone respond?
Does the platform act?
Does the law recognize what happened?
Can investigators identify the person responsible?
Can the victim continue going to school or work without being told that nothing can be done because the image was “only AI”?
Those questions determine whether legislation means anything.
Technology has always forced law to chase behavior.
The printing press.
Photography.
Telephones.
Broadcasting.
The internet.
Social media.
Each innovation created extraordinary benefits and new forms of harm.
Artificial intelligence is repeating that pattern at extraordinary speed.
But synthetic sexual imagery presents an especially intimate challenge because it attacks something law has historically assumed was difficult to counterfeit convincingly:
Embodied identity.
Your face used to provide some evidence that you were somewhere.
Now it can be detached from you.
Your likeness can perform actions you never performed.
Your synthetic double can travel places you never went.
That means societies need new rules for an old principle.
You should have meaningful control over whether your identity is used to depict you in intimate sexual material.
That principle need not require choosing between dignity and freedom.
The challenge is protecting both with precision.
A democratic society should be capable of saying two things simultaneously:
Free expression deserves powerful protection.
And manufacturing or distributing sexually explicit depictions of people without their consent can cause profound harm that deserves a serious legal response.
Those principles are not automatically enemies.
The hard work is drawing the boundary correctly.
The overwhelming political support surrounding the TAKE IT DOWN Act suggested lawmakers believed the boundary could no longer be postponed.
AI had changed the threat faster than older laws had changed the protection.
Victims were already living with the consequences.
The question was no longer whether synthetic media might eventually become dangerous.
The danger had arrived.
Now comes the less dramatic work.
Implementation.
Enforcement.
Safeguards.
Court interpretation.
Platform compliance.
Public education.
And constant adjustment as the technology evolves.
Because AI will not stop improving while lawmakers debate it.
Tomorrow’s deepfakes will likely be easier to produce and harder to distinguish from authentic media than yesterday’s.
That makes the principle established today increasingly important.
A person’s dignity cannot depend on whether strangers can determine that an explicit image is fake.
The violation begins earlier.
It begins when someone’s identity is taken and sexualized without permission.
The technology may be artificial.
The humiliation isn’t.
The pixels may be synthetic.
The fear isn’t.
The depicted body may never have existed.
The person crying over the image does.
And whatever laws ultimately govern the AI era, that human being cannot become the detail we forget while arguing about the technology.
