President Trump has now signed the Take It Down Act, criminalizing sexual deepfakes at a federal stage within the US. On the similar time, the CivitAI group’s bid to ‘clear up its act’ relating to NSFW AI and celeb output has finally didn’t appease cost processors, main the positioning to hunt options or face shutdown. All this within the mere two weeks because the oldest and largest deepfake porn website on the earth went offline…
It has been a momentous few weeks for the state of unregulated picture and video deepfaking. Simply over two weeks in the past, the quantity #1 area for the group sharing of superstar deepfake porn. Mr. Deepfakes, out of the blue took itself offline after greater than seven years in a dominant and much-studied place as the worldwide locus for sexualized AI superstar content material. By the point it went down, the positioning was receiving a median of greater than 5 million visits a month.
Background, the Mr. Deepfakes area in early Could; inset, the suspension discover, now changed by a 404 error, because the area was apparently bought by an unknown purchaser on the 4th of Could, 2025 (https://www.whois.com/whois/mrdeepfakes.com). Supply: mrdeepfakes.com
The cessation of companies for Mr. Deepfakes was formally attributed to the withdrawal of a ‘vital service supplier’ (see inset picture above, which was changed by area failure inside every week). Nonetheless, a collaborative journalistic investigation had de-anonymized a key determine behind Mr.Deepfakes instantly previous to the shutdown, permitting for the chance that the positioning was shuttered for that particular person’s private and/or authorized causes.
Across the similar time, CivitAI, the industrial platform broadly used for superstar and NSFW LoRAs, imposed a set of surprising and controversial self-censorship measures. These affected deepfake era, mannequin internet hosting, and a broader slate of latest guidelines and restrictions, together with full bans on sure marginal NSFW fetishes. and what it termed ‘extremist ideologies’.
These measures had been prompted by cost suppliers apparently threatening to withdraw companies from the area except modifications relating to NSFW content material and superstar AI depictions had been made.
CivitAI Lower Off
As of right this moment, it seems that the measures taken by CivitAI haven’t appeased VISA and Mastercard: a brand new submit† on the website, from Group Engagement Supervisor Alasdair Nicoll, reveals that card funds for CivitAI (whose ‘buzz’ digital cash system is usually powered by real-world credit score and debit playing cards) will likely be halted from this Friday (Could twenty third, 2025).
It will stop customers from renewing month-to-month memberships or shopping for new buzz. Although Nicoll advises that customers can keep present membership privileges by switching to an annual membership (costing†† $100-$550 USD) earlier than Friday, clearly the long run is considerably unsure for the area right now (It needs to be famous that annual memberships went stay on the similar time that the announcement in regards to the lack of cost processors was made).
Relating to the dearth of a cost processor, Nicoll says ‘We’re speaking to each supplier snug with AI innovation’.
As to the failure of current efforts to adequately rethink the positioning’s oft-criticized insurance policies round celeb AI and NSFW content material, Nicoll states within the submit:
‘Some cost firms label generative-AI platforms excessive danger, particularly once we permit user-generated mature content material, even when it’s authorized and moderated. That coverage selection, not something customers did, compelled the cutoff.’
A remark from consumer ‘Faeia’, designated as the corporate’s chief of employees of their CivitAI profile*, provides context to this announcement:
‘Simply to make clear, we’re being faraway from the cost processor as a result of we selected to not take away NSFW and grownup content material from the platform. We stay dedicated to supporting every kind of creators and are engaged on various options.’
As a conventional driver of latest applied sciences, it isn’t unusual for NSFW content material for use to kick-start curiosity in a website, expertise or platform – just for the preliminary adherents to be rejected as soon as sufficient ‘reputable’ capital and/or a user-base is established (i.e., sufficient customers for the entity to outlive, when shorn of a NSFW context).
It appeared for some time that CivitAI would observe Tumblr and various different initiatives down this route in direction of a ‘sanitized’ product able to overlook its roots. Nonetheless, the extra and rising controversy/stigma round AI-generated content material of any variety represents a cumulative weight that appears set to stop a last-minute rescue, on this case. Within the meantime, the official announcement advises customers to undertake crypto instead cost technique.
Pretend Out
The arrival of President Donald Trump enthusiastically signing the Federal TAKE IT DOWN Act is more likely to have influenced a few of these occasions. The brand new legislation criminalizes the distribution of non-consensual intimate imagery, together with AI-generated deepfakes.
The laws mandates that platforms take away flagged content material inside 48 hours, with enforcement overseen by the Federal Commerce Fee. The prison provisions of the legislation take impact instantly, permitting for the prosecution of people who knowingly publish or threaten to publish non-consensual intimate photographs (together with AI-generated deepfakes) throughout the purview of america.
Whereas the legislation obtained uncommon bipartisan help, in addition to backing from tech firms and advocacy teams, critics argue it might suppress reputable content material and threaten privateness instruments like encryption. Final month the Digital Frontier Basis (EFF) declared opposition to the invoice, asserting that the takedown mechanisms it mandates goal a broader swathe of fabric than the narrower definition of non-consensual intimate imagery discovered elsewhere within the laws.
‘The takedown provision in TAKE IT DOWN applies to a wider class of content material—probably any photographs involving intimate or sexual content material—than the narrower NCII definitions discovered elsewhere within the invoice. The takedown provision additionally lacks vital safeguards in opposition to frivolous or bad-faith takedown requests.
‘Providers will depend on automated filters, that are infamously blunt instruments. They incessantly flag authorized content material, from fair-use commentary to information reporting. The legislation’s tight time-frame requires that apps and web sites take away speech inside 48 hours, not often sufficient time to confirm whether or not the speech is definitely unlawful.
‘Because of this, on-line service suppliers, significantly smaller ones, will possible select to keep away from the onerous authorized danger by merely depublishing the speech fairly than even trying to confirm it.’
Platforms now have as much as one 12 months from the legislation’s enactment to determine a proper notice-and-takedown course of, enabling affected people or their representatives to invoke the statute in looking for content material elimination.
Because of this though the prison provisions are instantly in impact, platforms are usually not legally obligated to adjust to the takedown infrastructure (similar to receiving and processing requests) till that one-year window has elapsed.
Does the TAKE IT DOWN Act Cowl AI-Generated Superstar Content material?
Although the TAKE IT DOWN Act crosses all state borders, it doesn’t essentially outlaw all AI-driven media of celebrities. The act criminalizes the distribution of non-consensual intimate photographs, together with AI-generated deepfakes, solely when the depicted particular person had a affordable expectation of privateness:
The act states:
“(2) OFFENSE INVOLVING AUTHENTIC INTIMATE VISUAL DEPICTIONS.—
“(A) INVOLVING ADULTS.—Besides [for evidentiary, reporting purposes, etc.], it shall be illegal for any particular person, in interstate or international commerce, to make use of an interactive laptop service to knowingly publish an intimate visible depiction of an identifiable particular person who just isn’t a minor if—
“(i) the intimate visible depiction was obtained or created below circumstances wherein the particular person knew or fairly ought to have identified the identifiable particular person had an inexpensive expectation of privateness;
“(ii) what’s depicted was not voluntarily uncovered by the identifiable particular person in a public or industrial setting [i.e., self-published porn];
“(iii) what’s depicted just isn’t a matter of public concern; and
“(iv) publication of the intimate visible depiction—
“(I) is meant to trigger hurt; or
“(II) causes hurt, together with psychological, monetary, or reputational hurt, to the identifiable particular person.
The ‘affordable expectation of privateness’ contingency utilized right here has not historically favored the rights of celebrities. Relying on the case legislation that ultimately emerges, it is attainable that even specific AI-generated content material involving public figures in public or industrial settings might not fall below the Act’s prohibitions.
The ultimate clause about figuring out the extent of hurt is famously elastic in authorized phrases, and on this sense provides nothing significantly novel to the legislative burden. Nonetheless, the intent to trigger hurt would appear to restrict the scope of the Act to the context of ‘revenge porn’, the place an (unknown) ex-partner publishes actual or pretend media content material of an (equally unknown) different ex-partner.
Whereas the legislation’s ‘hurt’ requirement could appear ill-suited to instances the place nameless customers submit AI-generated depictions of celebrities, it may show extra related in stalking eventualities, the place a broader sample of harassment helps the conclusion that a person has intentionally and maliciously focused a public determine throughout a number of fronts.
Although the Act’s reference to ‘lined platforms’ excludes non-public channels similar to Sign or electronic mail from its takedown provisions, this exclusion applies solely to the duty to implement a proper elimination mechanism by Could 2026. It doesn’t imply that non-consensual AI or actual depictions shared via non-public communications fall exterior the scope of the legislation’s prison prohibitions.
Clearly, an absence of on-site reporting mechanisms doesn’t hinder affected events from reporting what’s now unlawful content material to the police; neither are such events precluded from utilizing no matter typical contact strategies a website might make accessible to make a criticism and request the elimination of offending materials.
The Rights Left Behind
Greater than seven years of mounting public and media criticism over deepfake content material seem to have culminated inside an unusually quick span of time. Nonetheless, whereas the TAKE IT DOWN Act provides sweeping federal prohibitions, it might not apply in each case involving AI-generated simulations, leaving sure eventualities to be addressed below the rising patchwork of state-level deepfake laws, the place the legal guidelines handed usually mirror ‘native curiosity’.
As an example, in California, the California Celebrities Rights Act limits the unique use of a star’s identification to themselves and their property, even after their dying; conversely, Tennessee’s ELVIS Act focuses on safeguarding musicians from unauthorized AI-generated voice and picture reproductions, with every case reflecting a focused method to curiosity teams which are distinguished at state stage.
Most states now have legal guidelines concentrating on sexual deepfakes, although many cease in need of clarifying whether or not these protections prolong equally to personal people and public figures. In the meantime, the political deepfakes that reportedly helped spur Donald Trump’s help for the brand new federal legislation might, in observe, run up in opposition to constitutional limitations in sure contexts.
† Archived model: https://internet.archive.org/internet/20250520024834/https://civitai.com/articles/14945
†† Archived model (doesn’t characteristic month-to-month costs): https://internet.archive.org/internet/20250425020325/https://civitai.inexperienced/pricing
* The precise ‘chief of employees’ to the CEO at CivitAI is listed at LinkedIn below an unrelated identify, whereas the similar-sounding ‘Faiona’ is an official CivitAI employees moderator on the area’s subreddit.
First revealed Tuesday, Could 20, 2025
