Congress is crafting new legislation to go after perpetrators of deepfakes. It’s a great start, but it focuses only on people making and sharing non-consensual deepfake porn; it completely ignores those companies at the root of the problem: those that enable the creation and proliferation of deefpakes, and it shifts the burden to victims to find and go after their perpetrators, after the harm has been caused.

To put a real stop to the harm we need to look at the root of the problem: tech companies like Microsoft, Facebook, Twitter/X, and Google that have built, and continue to manage and profit from the technology that allows for the easy creation and distribution of deepfake porn. Even as it’s become clear these companies are perpetuating harm, they have failed to take adequate steps to stop it. Worse, many have actually pulled back safeguards and disbanded teams that would help curb the problem

Now that deepfakes have become a major media story, you have Satya Nadella, CEO of Microsoft, saying “we have to act,” while not specifying or committing to real action—and with most other tech leaders remaining completely silent.

There are good conversations in the AI safety community about what the best approaches are for reining in deepfakes, but until the tech companies at the root of the problem act, it will be meaningless.

Some parts of solving the problem are harder than others, but there is plenty that can and should be done right now:

* AI companies (like Microsoft) should stop releasing software that has been shown to create harmful, non-consensual deepfakes, until they can prove that it is safe.

* Social media platforms (like Facebook and Twitter/X) should take much stronger steps to detect deepfakes; freeze accounts that appear to have distributed harmful, non-consensual deepfakes; and permanently ban those that have been determined to have done so.

* Cloud providers (like Amazon) should drop large websites that are clearly and overtly in the business of creating and distributing non-consensual deepfakes.

Congress is now working on The DEFIANCE Act of 2024, which would make it a crime to produce and distribute non-consensual deepfake images, audio, and video–an important part of what’s needed and which UltraViolet has recently started a campaign to support.

But Congress has yet to hold accountable the tech companies that are at the core of the problem, and that control the means of production and distribution of deepfakes. These companies have poured billions into the technology that makes deepfake creation and sharing possible. It’s time they prioritize addressing the harms they’ve created as a result, and invest in making such technology safe from causing harm.

Until they can demonstrate that a baseline level of safety has been achieved, with no one in the future being subjected to what Taylor Swift and others currently are experiencing, these companies need to do all that’s in their power to stop the harm created by the spread of non-consensual deepfakes.