What changed, and when
Platforms and lawmakers change the rules for AI personas roughly every three months. Most people only notice when their reach drops or a notice arrives. This page lists each change with its date, what it means for you and the source. Newest first.
We never rewrite an entry. If one turns out to be wrong, we add a correction with its own date, so you can always see what we changed.
Last updated · sources last checked · Follow by RSS
Show changes for
Snapchat’s updated monetisation policy took effect: AI content earns only when disclosed
The Creator Monetization Policy in effect from 21 September 2026 says content made with AI tools can be monetized if it is original, not misleading, and the use of AI is disclosed somewhere in the content or the creator profile. Deceptively manipulated AI media is not monetizable. Say it is AI in the profile and in the content before you count on Snapchat income.
Meta launched Meta One, with clickable links in Instagram reels from $49.99 a month
Meta launched Meta One, a paid subscription for Instagram, Facebook and WhatsApp. The creator plans start at $14.99 a month (Essential, with the verified badge) and $49.99 a month (Advanced, with clickable links in up to 4 posts and 4 reels a month and shared access without a password). The badge needs a passed identity check, and the verified profile name must match the legal name on your ID. So an AI persona cannot carry the badge under its own name. The other features do not depend on the check.
Source: Meta Newsroom — Introducing Meta One
What to do nowThe "AI creator" label became "AI-generated profile"
If the person your account shows is AI-generated, Instagram expects the account to carry the new label. Accounts without it are shown less to non-followers in recommendations. A growing account depends on that traffic. Using AI only as a tool for editing or captions is not covered.
Source: Instagram Creators — AI-generated profile label
What to do nowCorrection,
The reach claim above was too broad. Meta’s own wording is narrower: accounts it believes are AI-generated profiles and have not labelled themselves "may see limits to their account’s reach by becoming ineligible to appear in recommendations until they add the profile label". So the account is taken out of recommendations completely. It is not just shown a little less. This only affects accounts Meta has identified. Meta tells you first, and you can appeal in Account Status. Meta also says that using the label "will neither help nor hurt your account’s reach". One more point: the post label is required only for photorealistic video or realistic audio. Meta does not ask you to label AI-generated images. So if your persona posts only still images, the account label is the one you need.
An OnlyFans owner said AI will “never” replace its creators
James Sagan’s company Architect Capital bought a 16% stake in OnlyFans in May 2026. He said the platform “will never use AI in any capacity to disrupt creators”. It would never use AI to cut creators out, and never to automate chats. The same report says OnlyFans has banned fully AI-generated accounts. This is the second public statement in ten months that says the same thing. OnlyFans’ own terms only ask for an #ai caption on AI content.
Source: Tubefilter — OnlyFans investor on AI and creator disintermediation
What to do now Affiliate linksX closed Creator Revenue Sharing and replaced it with Original Content Rewards
X stopped taking new members into Creator Revenue Sharing on 7 August 2026 and closed it on 7 September. The new Original Content Rewards programme opened to existing members from 8 September. It asks for a Premium subscription, 500 verified followers and 500,000 Home Timeline impressions from verified users in 90 days. It pays only for original content, and content "created or posted using automated means" does not count. X has no other AI rule for the programme apart from a 90-day suspension for undisclosed AI videos of armed conflict.
Source: X Help — Original Content Rewards
What to do nowArticle 50 transparency obligations started applying
If you publish photorealistic AI-generated content that shows people, you must say clearly that it is artificially generated. People must be told at the latest when they first see it. This applies whether or not a real person is shown, and whatever your intention is. Deployers (the people who use an AI tool, such as you) also must not remove the provenance markers that a generator puts into the file.
Source: EU AI Act Art. 50(1)
What to do nowCorrection,
The last sentence above was wrong. Article 50 does not say that you must not remove the machine-readable markers. Article 50(2) makes the providers of AI generators mark their output. It sets no rule for deployers about those markers. The only ban on removing them on purpose is in the voluntary Code of Practice on AI-generated content, as a clause that signatory providers put into their terms. Keep the markers anyway: platforms read them to label your posts automatically.
California’s AI Transparency Act now applies to the tool makers
SB 942 now applies to generative AI providers with more than one million monthly users. They must put a hidden (latent) disclosure into AI-generated images, video and audio, offer the option of a visible one, and provide a free public detection tool. Large online platforms follow from 1 January 2027. From then they must detect that provenance data and show it. Makers of capture devices such as cameras follow from 2028. None of this makes you file or label anything when you post content. What changes for you: the tools add watermarks, and the platforms get better at reading them.
Source: Morgan Lewis — New California AI disclosure rules become operative
What to do nowSnapchat stopped recommending wholly AI-generated videos on Spotlight
Snap said that from July 2026, wholly AI-generated videos are no longer eligible for recommendation on Spotlight. Content edited with Snapchat’s own AI tools stays eligible and carries transparency indicators. For a fully AI persona this closes the main way new people find you on Snapchat. Plan its reach on another platform.
Source: Snap Newsroom — Rewarding Authentic Creativity on Spotlight
Commission guidelines read the editorial exception narrowly
The guidelines made two things clear. You have to label even if you do not mean to deceive anyone, and even if no real person is shown. And the exception for AI-assisted text that a human has reviewed only applies when someone takes real editorial responsibility. A quick look does not count.
Source: European Commission guidelines on Article 50 transparency obligations
What to do nowReddit said it now uses LLMs to catch coordinated fake behaviour
Reddit reported that it blocks 23 million spam views a day. It said it uses LLMs (AI language models) to catch “the highly subtle, coordinated patterns of fake behavior” that earlier systems missed, and that it checks signals from the moment an account is created. For a persona operation on Reddit, the real risk is a group of accounts that act together. A self-promotion ratio is not the risk.
Source: TechCrunch — Reddit is using LLMs to solve a problem LLMs largely created
What to do nowThe Council adopted the AI Omnibus: new bans, and 2 December for marking
The Council gave its final approval to the Digital Omnibus on AI. Two parts matter if you run a persona. First, providers of generative AI tools get a shorter grace period to build machine-readable marking of their output into the tool. The deadline is 2 December 2026. That is a rule for the tool maker. Your own rule as a deployer, to label deepfakes, has applied since 2 August 2026 and did not move. Second, the Act gets a new ban on AI practices that generate non-consensual sexual and intimate content or child sexual abuse material.
What to do nowA federal AI labelling bill was introduced again
Senators Schatz, Curtis and Warner introduced the AI Labeling Act of 2026. It would make generative AI providers attach a visible label and a machine-readable record. It would also make large platforms flag such content and not strip the marking. A first version was introduced in 2023 and did not pass. It is listed here because some pages quote US fines and enforcement dates from this bill as if it were law. It is not law.
Source: Music Business Worldwide — US senators revive bill to force AI content labels
What to do nowThe final Code of Practice on labelling AI-generated content was published
The code explains how providers and deployers of generative AI can follow the Article 50 transparency rules. A first draft came in December 2025 and a second in March 2026. The code is voluntary, and the rules themselves are still those of Article 50. But it is the Commission’s own worked example of what correct marking and labelling looks like. Nobody had that in writing before.
Source: European Commission — Code of Practice on transparency of AI-generated content
What to do nowNew York made ads say when they show an AI-generated person
Section 396-b of New York’s General Business Law took effect. Whoever produces or creates an ad for a business must say clearly in the ad that it contains a "synthetic performer": a person made with generative AI or software who is not a recognisable real performer. An AI persona in a brand deal fits that description. The fine is $1,000 for a first violation and $5,000 for each one after. Audio ads and AI used only to translate a real performer are excluded, and the media that publish the ad are not liable. The law gives no wording, only that the note must be conspicuous.
Source: NY Senate — General Business Law § 396-b
What to do nowReddit’s manipulated-content policy, as updated: one line on AI
As of this update, Reddit allows AI-generated content, subject to each community’s rules. Across the whole platform it bans only AI content that “presents itself as human-generated”. There is no AI label, toggle or flag. You say it is AI in the post itself. Whether a subreddit allows AI content at all is decided by its moderators, and you find that in its sidebar.
Source: Reddit Help — Manipulated Content and Misleading Behavior
What to do nowThe AI Omnibus deal kept Article 50 on schedule, with one grace period
The political agreement on the Digital Omnibus postponed the deadlines for high-risk AI. It left the Article 50 transparency rules in place, applying from 2 August 2026. There is one exception. Generative AI systems that were already on the market before that date get until 2 December 2026 for the machine-readable marking required by Article 50(2). That is a rule for the provider, built into the generator. The agreement did not move the transparency rules themselves.
Source: Gibson Dunn — EU AI Act Omnibus agreement and key changes
What to do nowThe 85% founding earning rate ended
The 85% rate offered to creators who joined in 2022 ended. Those accounts moved to the standard 80%. Nothing about the accounts changed, but a fifth of every dollar no longer goes to the creator. It is the clearest example on this page that a payout rate you plan with can change.
Source: Fanvue Help — Earning Rate Changes for Creators on Fanvue
What to do now Affiliate linksX added a "Made with AI" label that creators set themselves
A switch in the post composer lets you mark a post as containing media made or changed with AI. It was voluntary at launch, and X published no penalty for leaving it off. X did not publish a help-centre article about it. How and where it applies is known from news reports and not from X.
Source: Techweez — X rolls out “Made with AI” label
What to do nowAI voice calls announced, not yet live
Fanvue announced a feature for subscribers: a fan can talk to an AI version of the creator, in the creator’s own cloned or designed voice. The app shows whether a call is AI or live. Fanvue says the feature "will soon be coming". What exists today is the AI Self setup and AI voice notes. Do not plan your work around the calls until they are live.
Source: Fanvue Blog — AI Voice Calls and the Future of the Creator-Fan Connection
India made platforms label AI-generated media and ask users to declare it
India changed its IT Rules. The changes were published on 10 February and have applied since 20 February 2026. Platforms must put a clear label on synthetic (AI-generated) audio and visual content, and must not let anyone remove that label. Large social media platforms must also ask users whether an upload is synthetic and check the answer with their own technical tools. A draft rule that the label must cover 10% of the content was dropped. The time to take content down was cut to three hours, and to two hours for intimate deepfakes.
Source: Freshfields — India targets deepfakes and AI-generated content (IT Rules 2026)
What to do nowFirst draft of the EU code on marking and labelling AI content
The Commission published the first draft of a voluntary Code of Practice on marking and labelling AI-generated content. It was the first attempt to describe what following Article 50 looks like in practice, beyond the text of the Act. A second draft followed on 3 March 2026 and the final code on 10 June 2026.
Source: European Commission — Code of Practice on transparency of AI-generated content
What to do nowTikTok let viewers turn AI content down, and added invisible watermarks
TikTok began testing an AI-generated content control in Manage Topics. It lets people choose how much AI content appears in their For You feed. TikTok also added invisible watermarks, readable only by TikTok, to content made with its own AI tools and to uploads that carry C2PA Content Credentials. It reported that it had labelled more than 1.3 billion videos. For a persona account the control is the new part. Some viewers can now turn down AI content as a whole, labelled or not.
Source: TikTok Newsroom — More ways to spot, shape and understand AI-generated content
What to do nowOnlyFans’ CEO said there are no “wholly AI” accounts on the platform
Speaking at Bloomberg Tech London, CEO Keily Blair said OnlyFans has no “wholly AI” accounts or tools. She also said AI cannot replace human-created content there. OnlyFans’ Terms of Service ask for AI content to be captioned with a tag such as #ai. They say nothing we could find about fully AI accounts, so statements like this one are the clearest evidence on that question.
Source: Investing.com — OnlyFans CEO comments on AI-generated content, Bloomberg Tech London
What to do now Affiliate linksPinterest let people see less AI content in beauty, art, fashion and home decor
Pinterest’s updated tuner lets people see less GenAI content on image Pins in categories prone to AI, such as beauty, art, fashion and home decor. On 12 December 2025 it added food and drink. Pinterest also said its labels would become more noticeable. A lifestyle persona posts in exactly these categories, so part of its audience can now turn it down.
Source: Pinterest Newsroom — New tools to give users more control over GenAI content
California pushed its AI Transparency Act back to August 2026
AB 853 became law. It delays the California AI Transparency Act from 1 January 2026 to 2 August 2026. That is the same day the EU’s Article 50 rules began to apply.
Source: CalMatters Digital Democracy — AB 853 California AI Transparency Act
What to do nowTikTok rewrote its deepfake rule, effective 13 September 2025
The guideline update widened the rule. Before, it covered content that “falsely shows public figures in certain contexts”. Now it covers content “misleading about matters of public importance or harmful to individuals”. The update also removed the wording on AI endorsements. The rule is not aimed at persona accounts, and the label requirement did not change. It does show that TikTok was still rewriting its AI section two and a half years after it started.
Source: TechCrunch — TikTok's new guidelines add subtle changes for LIVE creators, AI content
What to do nowObligations for general-purpose AI model providers started applying
The rules for providers of general-purpose AI models took effect. So did the Act’s governance rules and the naming of national authorities. Models already on the market have until 2 August 2027. None of this applies to you as an operator. It applies to the companies whose models power the image and video tools you use.
Source: EU AI Act implementation timeline
What to do nowFacebook cut pay and reach for accounts that reuse other people’s content
Facebook said accounts that repeatedly reuse someone else’s videos, photos or text posts lose access to its monetisation programmes for a period of time and get reduced distribution on everything they share. A persona that posts its own generated content is not the target. Reposting other creators’ clips or trends unchanged is.
Source: Facebook for Creators — Combating unoriginal content
Pinterest started labelling AI Pins “AI modified” worldwide
Pinterest rolled out its Gen AI labels globally. Image Pins that its systems detect as AI-generated or AI-modified show “AI modified” in close-up. Pinterest reads IPTC metadata and also runs its own classifiers, so a label can appear without any marker in the file. It set up an appeal for creators who think a label is wrong.
The AI Act’s prohibitions and AI literacy requirement started applying
These were the first parts of the Act to apply: the list of banned AI practices and the AI literacy requirement. The bans on non-consensual intimate imagery and child sexual abuse material were split off and apply later, from 2 December 2026.
Source: EU AI Act implementation timeline
What to do nowCorrection,
The last sentence above was wrong. The bans on AI that generates non-consensual sexual and intimate content or child sexual abuse material were not split off from the 2025 list. They are a new provision. The Digital Omnibus, which the Council adopted on 29 June 2026, added them to the Act. They apply from 2 December 2026. See that entry: its source is the Council’s own press release and not a third-party timeline.
California’s AI Transparency Act (SB 942) was signed
The Governor approved the bill. It sets rules for generative AI providers with more than one million monthly users: a disclosure embedded in the file, an optional visible one, and a free detection tool. It sets no rule for individual users or for people who post content. It was first due to start on 1 January 2026 and was later delayed.
Source: California Legislative Information — SB 942 AI Transparency Act
What to do nowThe AI Act entered into force, starting the clock
Nothing changed for you on this date. It is listed because the labelling rules that started applying two years later were already scheduled from this day. So the labelling rules did not come suddenly: the date was known two years before.
Source: EU AI Act implementation timeline
What to do now"Made with AI" was renamed "AI info" after it mislabelled real photos
Meta renamed the label in all its apps after photographers found that normal retouching triggered it. The detection did not change. It still reads C2PA and IPTC metadata. The problem was the old name, which said more than the metadata could prove. The same metadata decides whether a generated image gets a label at all.
Source: Meta Newsroom — Our Approach to Labeling AI-Generated Content and Manipulated Media
What to do nowTikTok started auto-labelling AI content from C2PA metadata
TikTok became the first video platform to read C2PA Content Credentials and label AI-generated content automatically. That includes content made in other tools. TikTok also attaches Credentials to content made on TikTok, so they stay in the file after a download. This shows why provenance metadata matters: on TikTok it decides whether your post gets a label you did not set yourself.
Source: TikTok Newsroom — Partnering with our industry to advance AI transparency and literacy
What to do nowSnapchat announced a watermark for images made with its own AI tools
Snap said images made with its generative AI tools would get a small ghost logo with a sparkle icon when they are exported or saved to the camera roll. The watermark only covers Snap’s own tools. Images made elsewhere get no automatic label on Snapchat, so saying it is AI is up to you.
Source: Snap Values — AI on Snapchat: Improved Transparency, Safety, and Policies
Meta announced the first "Made with AI" labels
Meta announced labels for a wider range of video, audio and image content. They are applied when Meta detects industry-standard AI indicators, and they appeared on organic (unpaid) content from May 2024. These were the first in a series of labels that ended with the AI-generated profile label on the account.
Source: Meta Newsroom — Our Approach to Labeling AI-Generated Content and Manipulated Media
What to do nowYouTube required creators to disclose realistic altered or synthetic content
YouTube started requiring a disclosure in Creator Studio for realistic content that a viewer could mistake for a real person, place, scene or event. Three things are exempt: clearly unrealistic content, beauty filters, and AI used only for scripts or captions. The label usually sits in the expanded description. For sensitive topics such as health, news, elections and finance, a more visible label appears on the video itself.
Source: YouTube Official Blog — Disclosing AI-generated content
What to do nowMeta said it would label AI images by reading C2PA and IPTC signals
Meta announced that it would label images on Facebook, Instagram and Threads when it detects industry-standard AI indicators in C2PA and IPTC metadata. It also added a disclosure tool that it would require for photorealistic video and realistic audio, with penalties for not using it. So Meta would detect still images itself, and the person posting would not have to declare them. That is the reason removing this metadata matters.
Source: Meta Newsroom — Labeling AI-Generated Images on Facebook, Instagram and Threads
What to do nowTikTok launched a label for AI-generated content
TikTok introduced a tool that lets creators label content “completely generated or significantly edited by AI”. It also said it would start testing automatic detection that adds the label without the creator doing anything. The label you set yourself came first. The automatic one arrived in May 2024.
Source: TikTok Newsroom — New labels for disclosing AI-generated content
What to do nowTikTok’s guidelines first required synthetic media to be disclosed
The new Community Guidelines said that synthetic or manipulated media showing realistic scenes must be “clearly disclosed”, with a sticker or caption such as “synthetic”, “fake”, “not real” or “altered”. The rule was aimed at very realistic depictions of real people and at edits that could mislead about real events. This is the earliest entry on this page. From the start the rule was about how realistic the content looks, and not about AI as such.
Source: TechCrunch — TikTok overhauls community guidelines, adds AI policy
What to do now