News
Ofcom Opens TikTok Age-Check Probe Under Online Safety Act
Ofcom has opened a formal TikTok investigation over age-check failures under the UK Online Safety Act, putting age inference methods at risk industrywide.
Ofcom opened a formal investigation into TikTok on July 16 over whether the platform’s age checks meet its child-protection duties under the UK’s Online Safety Act.
The probe targets “age inference.” That is the system TikTok uses to guess a user’s age from behavior. It does not rely on a verified ID or a face scan. Ofcom assigned the case a number, CW/01352/07/26. It falls under section 12 of the Online Safety Act 2023.
Age inference is the lowest-friction gate a social platform can run. No ID upload. No biometric scan. No drop-off at signup. TikTok is not alone in leaning on it. A ruling that it fails the law’s “highly effective” test would disqualify soft age assurance across the industry. Hard verification would then become the only compliant path for every major platform. That reshapes who can reach an under-16 audience. It also reshapes how branded content and TikTok Shop reach the youngest end of a creator’s following.
What Ofcom actually found
The investigation landed the same day as Ofcom’s first Age Assurance Report. The report covers the first six months after the children’s protection duties took effect in July 2025. It says Ofcom has already ruled out age inference as highly effective age assurance.
Ofcom tied that finding to TikTok. The regulator said the report suggests age inference models, such as those used by TikTok, may have failed to correctly identify a significant proportion of children, putting them at risk of exposure to harmful content.
Kate Davies, Ofcom’s group director for strategy and research, put it plainly on BBC Radio 4. “We have very serious questions about whether age inference can be highly effective,” she said.
TikTok rejects the premise. A spokesperson said the company strictly enforces age-appropriate experiences through expert-informed platform rules and advanced age inference technologies, in line with major industry peers. The spokesperson added that TikTok is “confident that we meet our Online Safety Act obligations and will work with Ofcom to demonstrate it.”
The stakes and the timeline
The penalty ceiling is steep. Ofcom can fine TikTok up to £18 million, about $24 million, or 10 percent of qualifying worldwide revenue, whichever is greater. TikTok publishes no standalone financials, and parent ByteDance keeps its accounts private. So the higher ceiling is an estimate, not a figure drawn from a primary disclosure. In the most serious cases, Ofcom can seek a court order. That order can require payment processors, advertisers, or ISPs to disrupt the platform’s UK business.
The first evidence-gathering stage runs about three months. Ofcom expects to publish an update in October 2026.
Ofcom told the wider industry to move now. It said services that use age inference should switch to other methods listed in its guidance as highly effective without delay.
The timing sharpens the pressure. The probe arrived a month after the UK government announced an under-16s ban from a range of social media platforms, with Prime Minister Keir Starmer preparing further legislation for a blanket ban. Ofcom plans to report to Parliament in late October on how over-16 age checks could work in practice, ahead of restrictions targeted for 2027.
The October update is the next fixed date. Parliament is still writing the ban the report says age inference cannot enforce.
News
TikTok Shop Swaps Strikes for Account Health Rating
TikTok Shop killed the strike-and-wait clock. A new 0-to-1,000 score now turns one bad livestream into compounding drag on a seller’s livelihood, and it hits affiliate creators too.
For years a TikTok Shop seller could take a policy strike and wait it out. As of July 2026 that clock is gone. TikTok Shop retired its Violation Points system this month. In its place sits a 0-to-1,000 Account Health Rating that follows a seller across a rolling 180-day window.
The change matters most for the people who make a living on TikTok’s shopping livestreams. Every seller starts at 200 points. They earn points back for fulfilled orders and passing policy quizzes. They lose points for violations scaled to severity. The old system reset every 90 days. The new one does not. One bad stream now creates compounding drag instead of a strike that expires.
The rule TikTok wrote in May finally bites
That architecture is what gives teeth to a rule TikTok wrote in May but could barely punish. The TechTimes analysis of the transition documents the connection. The May ban on AI-generated voices in shopping livestreams now feeds into the AHR. TikTok’s Prohibited Content guidelines tell hosts not to use “non-real-time verbal interaction such as AI-generated voices, audio recordings, or radio,” and instead to “engage directly with viewers using real-time verbal or sign language communication.”
The score declines in stages. Each stage costs a seller something specific. At 150 points, a seller cannot enroll in new mega campaigns and cannot create new listings for seven days. At 100 points, those blocks extend to 14 days. Livestream traffic drops, and products fall out of Shop Tab recommendations. Further restrictions hit at 50 points. At zero the account faces permanent deactivation.
Both the seller and the affiliate pay for one stream
The consequence does not stop at the merchant. A parallel Creator Health Rating applies the same logic to affiliate creators. It runs its own 0-to-1,000 scale from a 200-point start. TikTok rolled that system out to creators in January, ahead of the seller version.
So a single non-compliant AI-voice livestream can dent two accounts at once. The merchant’s AHR takes the hit. The affiliate creator’s CHR takes an independent one for the same broadcast.
That rewrites the risk math of the seller-affiliate relationship. An affiliate who drops below a 150 CHR loses eligibility to promote products in TikTok Shop’s 17 sensitive categories. TikTok runs that check in real time each time a creator selects a product. Many small sellers and affiliates lean on cheap production workflows: text-to-speech narration looped over static footage. That is now the exact behavior both scores penalize.
Continuous pressure on a $23 billion channel
The stakes scale with the channel. PYMNTS reported the eMarketer projection that TikTok Shop will reach $23.4 billion in US ecommerce sales in 2026, a 48% jump year over year that would place it ahead of Target, Costco and Best Buy by US ecommerce volume. TikTok is now imposing continuous compliance pressure on its fastest-growing storefront. That replaces a strike-and-reset model that let bad actors game the line.
There is one carve-out worth the reader’s attention. AI in pre-production stays legal. Scripting, editing, translation, and TikTok’s own Symphony creative tools for advertisers all remain allowed. The rule bars AI from standing in for a live human during the session itself. It does not stop AI from helping build the content before the camera turns on.
The AHR requirements document published June 25, less than a month before the switch flipped for sellers.
News
ExtraEmily’s Twitch Driving Ban and Reversal Split Streamers
A viral near-crash got ExtraEmily suspended, then reinstated within a day. Now Asmongold and other top streamers want Twitch to ban in-car streams for good.
A Mazda SUV honked, and that horn is the only reason ExtraEmily’s June 28 livestream did not end in a wreck.
Twitch suspended the IRL streamer, whose real name is Emily Xuechun Zhang, on June 30 after a clip of the near-collision went viral. The company reversed the ban roughly 24 hours later, according to Dexerto. She has about 978,000 followers on the platform.
The whipsaw exposed a gap Twitch’s IRL boom has been widening for years. The platform wrote rules against distracted driving but never barred streaming from a moving car outright, and its punishment here landed lighter than the offense. In-car broadcasting is a cheap, high-engagement format that the safety framework has not caught up to, and the exposure runs to viewers, the platform, and the streamer alike.
The near-miss came roughly one hour and 25 minutes into the June 28 broadcast, as Zhang made a left turn while appearing to look at her phone. The other driver’s horn alerted her. Neither vehicle was damaged.
“That was my bad,” she said on stream.
Zhang later explained she had been driving with her Tesla’s Autopilot on but disabled it shortly before the incident. Her channel then displayed a message citing a Community Guidelines violation, and she confirmed on Discord the suspension was for distracted driving.
The streamers pushing for a hard line
Twitch’s own rules already name the conduct. Its Self-Destructive Behavior policy states that dangerous or distracted driving that removes a streamer’s eyes from the road “is never acceptable on Twitch,” and reading chat behind the wheel is specifically prohibited.
What the rules do not do is bar in-car streaming outright, and the reversal convinced several top creators that the middle ground has failed. Zack “Asmongold” Hoyt posted his call for a full prohibition on X on June 30. “Ngl they need to just ban streaming and driving entirely. Want to stream yourself going somewhere? Get an Uber or have someone else drive. At best it is a neutral effect, at worst, it causes accidents and kills people. Pretty easy logical decision.”
Jesse Cox framed the risk as an existential one for the company, as reported by Tubefilter. “Every single person who streams and drives should be perma banned. Cars are dangerous machines and if someone dies, the lawsuit against Twitch will kill the site.”
DansGaming argued the rule used to be stricter. “I’m so confused, driving and streaming used to always be an instant ban because it’s so dangerous. When did this change?! Drivers are terrible and dangerous enough when they aren’t streaming. This should never be encouraged or allowed on Twitch.”
Why a blanket ban is harder than it sounds
An outright prohibition runs into the platform’s own economy. Tubefilter notes an entire Twitch category is dedicated to trucking, where a ban would strip working drivers of income, and that the high-stakes pull of streaming culture already nudges creators toward reckless driving.
For Zhang, the suspension marked her third driving-related ban. The first came in October 2023 for reading chat behind the wheel. The second came in April 2025 after she appeared to run a red light while looking at her phone on a livestream.
Zhang returned to air and addressed the incident, pledging to minimize in-car broadcasts and turn chat off when driving. “I want to learn from my mistake and do better in the future, so from now on I’m going to try to minimize the amount of driving I do on stream. And if I do drive on stream, I’m going to have the chat off. I just want to do better and not make the same mistake again and again. I do feel really bad. I’m learning from it, and I want to just do better.”
Twitch has not said whether it will codify a hard rule against streaming while driving. The category that made the clip possible is still live, and so is the trucking channel that any ban would have to reckon with.
News
YouTube Launches Four Gemini Creator Tools at Cannes Lions
YouTube put trend data, audience metrics, and creator discovery inside Google Ads at Cannes Lions, with no platform cut on direct deals. The layer it is absorbing is the one agencies have long billed for.
YouTube unveiled four Gemini-powered creator marketing tools at Cannes Lions on Tuesday, June 23, pushing trend data, brand metrics, and creator discovery deeper inside Google Ads.
Google detailed the tools in a company blog post. Insights Finder now carries more granular U.S. trending data on what is popular on YouTube in real time. Select Brand Pulse metrics now appear inside Insights Finder, so brands can read paid and organic presence from a single interface.
The other two aim at the people whose names sit inside those campaigns. A new Content and Creator Insights API hands agencies richer data on creators and audiences for media planning. Gemini will soon suggest which visuals lift Demand Gen campaign performance.
Compiling and interpreting that data has long been agency work, the billable layer between a brand and a creator. YouTube is moving that layer inside the ad platform where brands already spend. That shrinks what a broker can charge to sit in the middle of a deal.
The tools build on Creator Partnerships, the platform YouTube introduced at NewFronts in March as the successor to BrandConnect. YouTube said the system uses Gemini to give brands access to more than 3 million creators inside the YouTube Partner Program, a search that once meant spreadsheets or a talent roster. It lives inside YouTube Studio for creators and inside Google Ads and Display and Video 360 for advertisers.
For mega creators fielding constant inbound, the pitch is less friction and deals that sit beside existing analytics. For mid-tier and micro creators, the shift cuts deeper. Discovery by natural-language prompt surfaces channels a brand would never have found through a roster, and it lets creators compete on audience fit rather than name recognition.
The data already changes who gets found. YouTube said creators who opt in to share channel insights appear 60% more often in advertiser search results, a visibility gap that separates the discoverable from the invisible.
The paid side carries numbers too. Advertisers who promoted creator-led videos on Shorts saw an average 30% increase in conversion lift, YouTube reported, the figure it uses to sell brands on treating creator content as performance media rather than a side buy.
The trend data and Brand Pulse integration are live in the U.S. now. The Demand Gen creative recommendations arrive later, with no date set.
