Monthly summary
August 2026 summary: AI and child safety
Meta settled the largest US child-safety case against it in August. On 26 August, in the middle of a federal trial in Oakland, Meta agreed to pay up to $17.1 billion over ten years and to give teens default protections. Earlier in the month a New Mexico judge had ordered it to pay another $567 million. Regulators in Brazil, Australia and the Philippines acted against Discord and Roblox, and age assurance was again the remedy most often named.
Meta settled in the middle of a trial
On 6 August a New Mexico judge ordered Meta to pay $567 million into a youth mental-health abatement fund, on top of $375 million in earlier fines. The judge also ordered remedies that included a 90-hour monthly usage cap for under-18s and restrictions on AI chatbots. Meta said it will appeal.
A federal trial opened in Oakland on 18 August, with California, Colorado, Kentucky and New Jersey leading 29 states. The states alleged that Meta designed Facebook and Instagram to hook children and collected data from children under 13 without parental consent. A former Meta engineering director testified that the company took a “don’t ask, don’t tell” approach to under-13 users, and Instagram head Adam Mosseri testified that few teens used the Take a Break tool before it became a default. Meta denied the claims. The allegations are unproven.
On 26 August Meta settled with dozens of state attorneys general. It agreed to pay up to $17.1 billion over ten years and to give teens a two-hour daily limit, a midnight-to-6 a.m. block, hidden likes, a ban on cosmetic filters, stronger age assurance and an independent auditor. Texas settled separately for about $1 billion. Florida refused to join and said it would go to trial, and New Mexico stayed outside the deal. Critics, including former Meta insider Arturo Béjar and Fairplay, called it a milestone but said it left recommendation and infinite-scroll harms intact. Meta did not admit wrongdoing.
Cases against TikTok, Snap, Roblox and Discord multiplied
The Justice Department announced a $400 million COPPA settlement with TikTok and ByteDance on 21 August: $300 million at once and $100 million once a prior consent decree is vacated. There was no determination of liability, and Fairplay called the deal a disappointment. Pennsylvania’s attorney general sued TikTok on 11 August over its 13+ rating and safety-tool claims, and sued Snap on 25 August over addictive design and its age rating. Brazil’s data protection authority fined ByteDance R$153.7 million for processing children’s data without a valid legal basis. A Ninth Circuit panel ruled on 10 August that Section 230 is a defense to liability, not immunity from suit, which cleared the way for thousands of social media addiction suits to proceed. The allegations are unproven.
Senators and regulators targeted Roblox and Discord. Senators Josh Hawley and Dick Durbin opened a Senate Judiciary subcommittee investigation into Roblox on 12 August, citing 65,381 suspected child exploitation reports Roblox made to the National Center for Missing & Exploited Children in 2025, more than double 2024. A Texas judge ordered Roblox to produce documents by 15 November. Australia’s eSafety Commissioner secured a court-enforceable undertaking under which Roblox must make changes within three months and accept third-party audits. In Brazil, Discord suspended livestreams after the data protection agency ordered it to, following the July case of a 13-year-old girl who was allegedly coerced into suicide on the platform, and the federal government sued Discord for about $97 million. After a school shooting in Zamboanga, the Philippine interior secretary renewed a call to ban Roblox, although the South China Morning Post reported that officials had yet to examine the shooter’s devices. Roblox agreed to add the Philippine national ID to its age checks. The allegations are unproven.
Chatbot makers tightened teen controls and drew tests
OpenAI began rolling out ChatGPT for Teens on 18 August, with age prediction, default under-18 safeguards and parent alerts, and started age prediction in the EU on 25 August. Common Sense Media said the reliability of the parent alerts was unknown. Microsoft released a Copilot app with protections for users under 18, and Google turned on Gemini in Google Classroom for K-12 students of all ages, which drew concern from parent groups.
Two benchmarks found gaps, both with simulated users rather than real children. KORA rated 41 percent of 14,839 simulated conversations across seven AI apps as failing and found no significant safety gap between child and adult modes. apgard reported that about 30 percent of responses from 19 frontier models failed on sexual topics with young people. A JAMA Pediatrics study found that seeking emotional support from generative AI was a distinct marker of psychological distress in children and adolescents.
AI-generated abuse imagery reached the courts
Families sued xAI in federal court over Grok. A family alleged on 7 August that Grok was used to make abuse images of their 6-year-old daughter from authentic photos, and a proposed class action filed on 26 August alleged that xAI trained Grok on known child sexual abuse material. xAI sued the Bentonville photographer named in the Arkansas case; he has pleaded not guilty to criminal charges. The allegations are unproven.
Minnesota’s ban on nudification tools took effect on 1 August, and a federal judge refused xAI’s emergency request to block it. On 26 August the Seventh Circuit upheld the dismissal of a possession charge over AI-generated abuse images that depicted no real child, holding that the First Amendment protects private possession; production and distribution charges remained pending.
Reported numbers rose. Report Remove, run by the Internet Watch Foundation and the NSPCC, received 420 reports from UK children of faked explicit images of themselves in the first half of 2026, above its 397 for all of 2025. The National Center for Missing & Exploited Children said it made 448 reports involving AI-generated abuse material available to Georgia investigators in the first quarter of 2026, up from 14 in 2023. A Tech Transparency Project investigation reported by WIRED found more than 50 ads featuring AI-generated abuse imagery on Meta’s platforms between November 2025 and August 2026. The allegations are unproven.
Laws and age limits
In the US, the Senate Commerce Committee voted on 5 August to advance the Kids Online Safety Act and three bills on children and AI, and a federal judge refused to block California’s SB 976 on minors’ personalized feeds. New EU AI Act transparency rules took effect on 2 August, and a ban on AI systems that create non-consensual sexual deepfakes or child sexual abuse material applies from December.
Azerbaijan signed rules that will bar under-16s from social media accounts. New Zealand’s prime minister said his party would introduce a bill to ban under-16s, although coalition partner New Zealand First said it would not support it. Singapore’s prime minister said platforms will need reliable age checks instead of self-declared ages, and Kazakhstan’s president backed a draft under-16 ban. India’s IT ministry held a consultation that concluded a blanket ban for children would not be effective.
In Australia, ABC News reported that 81.5 percent of teens still used social media after the under-16 ban, down from 85.9 percent before it. Meta said it had removed access to more than 750,000 accounts, and a Senate committee recommended passing stronger eSafety powers. Microsoft began age checks for R18+ games on Xbox and the Microsoft Store, and Google made its Play Age Signals API available to developers worldwide.
What to watch
- Whether TikTok and YouTube match Meta’s teen measures; Meta urged both to.
- Florida’s trial against Meta, after the state refused the settlement.
- TikTok’s reply to the senators’ letter, due 1 September.
- Roblox’s changes under the Australian undertaking, due within three months.
- Roblox’s document production in Texas, due 15 November.
Sources (52)
- Meta reaches ~$17-18B settlement with 48 states over teen social-media addiction npr.org
- New Mexico judge orders Meta to pay another $567M in child-safety public nuisance case BBC News
- New Mexico judge orders Meta to pay $567M and impose child-safety measures Associated Press
- Trial opens in Oakland as 29 states accuse Meta of designing Instagram and Facebook to hook children Associated Press
- Former Meta engineer testifies Instagram took ’don’t ask, don’t tell’ stance on under-13s Associated Press
- Instagram chief Mosseri testifies that few teens used Take a Break before it became the default aljazeera.com
- Meta separately settles with Texas for ~$1B over similar child-safety allegations platformer.news
- Florida rejects Meta settlement as ’peanuts’ and vows trial while Texas strikes a separate $1 billion deal arstechnica.com
- Child-safety advocates say Meta’s new teen measures don’t go far enough and shift the burden onto parents Associated Press
- DOJ secures $400M COPPA settlement with TikTok and ByteDance justice.gov
- Fairplay condemns the DOJ-TikTok settlement as a ’sweetheart deal’ with no new protections fairplayforkids.org
- Pennsylvania attorney general sues TikTok over its 13+ rating and safety tool claims law.com
- Pennsylvania AG sues Snap over addictive design and misleading age rating news.bloomberglaw.com
- Brazil’s ANPD fines ByteDance R$153.7 million for mishandling children’s data gov.br
- 9th Circuit dismisses Meta and TikTok appeals, clearing social media addiction suits to proceed arstechnica.com
- Senate Judiciary subcommittee opens bipartisan investigation into child safety on Roblox hawley.senate.gov
- Texas judge orders Roblox to produce documents in state child-safety lawsuit star-telegram.com
- Australia’s eSafety regulator secures a court-enforceable Roblox undertaking on adult contact with children arstechnica.com
- Discord suspends livestreams in Brazil after data-protection regulator cites child-safety failures Associated Press
- Brazil’s federal government sues Discord for ~$97 million over failures to protect children aljazeera.com
- Philippine interior secretary renews call to ban Roblox after Zamboanga school shooting scmp.com
- Roblox agrees to add Philippine national ID to age checks and open a local office philstar.com
- OpenAI launched ChatGPT for Teens with age prediction and default under-18 safeguards cnn.com
- OpenAI begins rolling out its ChatGPT age-prediction system in the EU to auto-apply teen safeguards OpenAI
- Common Sense Media questions whether ChatGPT for Teens safeguards work CNBC
- Microsoft ships new Copilot app with dedicated ’Copilot for Young People’ protections support.microsoft.com
- Google expands Gemini in Classroom to K-12 students of all ages indianexpress.com
- KORA benchmark found no significant safety gap between child and adult modes of seven AI apps ETIH EdTech Innovation Hub
- apgard publishes benchmark for youth sexual safety in AI models apgard AI
- JAMA Pediatrics study links seeking emotional support from AI to distress in young people JAMA Network
- Federal lawsuit alleges xAI’s Grok was used to create child sexual abuse images of a 6-year-old nwaonline.com
- New federal class-action alleges xAI trained Grok’s image models on real child sexual abuse material mashable.com
- xAI sues Bentonville photographer accused of using Grok to make CSAM katv.com
- Court rejects xAI bid to block Minnesota’s AI nudification law ibtimes.co.uk
- Seventh Circuit upholds dismissal of private-possession charge over AI-generated child abuse images The Washington Post
- UK children reported a surge in explicit deepfakes of themselves, the IWF and NSPCC said The Guardian
- NCMEC gave Georgia investigators 448 AI-generated abuse reports in early 2026, up from 14 in 2023 Office of Senator Jon Ossoff
- Meta ran paid ads containing AI-generated CSAM across its platforms, researchers found WIRED
- Senate Commerce Committee advanced KOSA and several children’s AI safety bills US Senate Commerce Committee
- California judge refuses to block SB 976 on minors’ personalized feeds The Next Web
- EU AI transparency rules for chatbots took effect European Commission
- EU AI Act main obligations start 2 August with new bans on sexual deepfake tools elDiario.es (Europa Press)
- Azerbaijan signed rules that will bar under-16s from social media accounts APA (Azerbaijan state news agency)
- New Zealand proposes bill to ban under-16s from social media theguardian.com
- Singapore prime minister says social media platforms must add stronger child safeguards CNA (Channel NewsAsia)
- Kazakhstan’s president backs a social media ban for under-16s Qazinform
- Indian ministry consultation says a blanket social media ban for children would not be effective hindustantimes.com
- Australia’s under-16 social media ban shows limited effect eight months in ABC News (Australia)
- Meta reports removing 750,000+ under-16 accounts under Australia’s social-media ban Meta Newsroom
- Australian Senate backs stronger eSafety powers; Snapchat among five platforms under under-16-ban probe thenightly.com.au
- Microsoft begins age-verification checks for R18+ games on Xbox and Microsoft Store in Australia gadgetguy.com.au
- Google rolls out Play Age Signals API globally for age-appropriate app experiences biometricupdate.com
