On August 31, 2026, Instagram began testing tighter limits on undisclosed AI-generated profiles by renaming its “AI creator” label to “AI-generated profile” and reducing the reach of accounts that do not disclose AI-generated personas, officials said. The move aims to curb the spread of fake influencers by limiting their visibility in recommendation features like Explore and Reels unless properly labeled.

The updated policy, announced by Instagram on August 31, 2026, renames the existing “AI creator” profile label to “AI-generated profile” and introduces distribution limits for accounts featuring AI-generated personas that do not disclose their synthetic nature, officials said. Creators who apply the “AI-generated profile” label properly will not face penalties solely for using AI personas as their account identities, the company added.

Under the new rules, Instagram will reduce the reach of undisclosed AI-generated profiles in recommendation features such as Explore, Reels, and feed suggestions, limiting their visibility to users who do not already follow the account, according to Meta’s statements.

Meta clarified that the policy specifically targets profiles where the primary person is AI-generated, including virtual influencers, AI personas presented as real individuals, or AI faces used as recurring hosts. It does not apply to general AI-assisted content such as edited photos or AI-written captions, according to Meta’s guidelines. Accounts identified as featuring undisclosed AI-generated personas will receive notifications and can check their Account Status to view recommendation eligibility and either add the required label or appeal the classification to restore reach, sources confirmed.

This policy change applies primarily to Instagram, while Meta’s broader AI disclosure framework spans both Instagram and Facebook. Earlier efforts by Meta included the rollout of AI content labels such as “AI info” and “Made with AI” across both platforms to identify AI-generated video, audio, and images based on detection signals and self-disclosure. However, the new “AI-generated profile” label is unique to Instagram and focuses on the identity of the account rather than individual content pieces, according to Meta’s official communications. Facebook continues to follow general AI labeling rules for content but does not have a widely reported profile-level label equivalent to Instagram’s new designation.

Meta employs a combination of automatic detection methods—including industry-standard signals like Coalition for Content Provenance and Authenticity (C2PA) metadata embedded in files—and manual disclosure tools to identify AI-generated content and personas, according to technical documentation and company statements. Instagram will identify accounts featuring AI-generated people without the proper label and limit their visibility in recommendations until the label is applied or an appeal is successful. Posts from unlabeled AI persona accounts may be excluded from Explore, Reels recommendations, and feed suggestions for non-followers, directly affecting audience growth and monetization potential, officials said.

Creators can restore their reach by proactively applying the “AI-generated profile” label in their profile settings or by appealing if they believe their account was mistakenly flagged as AI-generated. Meta acknowledged that detection is not perfect and allows for human review in appeals. For AI-generated or AI-altered advertisements, Meta already enforces stricter rules, including ad rejection, forced labeling, reduced distribution, or removal if AI content is not properly disclosed, according to Meta’s advertising policies.

The introduction of tighter limits on undisclosed AI-generated profiles follows growing complaints from human creators who say fake AI personas compete for reach and brand deals without transparent labeling. Meta’s guidance advises creators to clearly disclose AI personas as “not a real person,” often through explicit text in bios such as “Virtual character… Not a real person,” reflecting concerns about misleading audiences, according to industry compliance guides and creator community reports. The lack of disclosure has raised trust issues and may undermine the negotiating power of human influencers with brands, sources said.

Meta’s prior rollout of “Made with AI” labels and “AI info” functionality between 2024 and 2026 was partly a response to user and regulator concerns over deepfakes and misleading AI media. The new profile-level rule represents a more direct approach to addressing fake influencers as account identities, according to Meta’s public statements. Industry analyses indicate that brands have experimented heavily with AI influencers who do not disclose their synthetic nature, prompting calls from creator communities for mandatory transparency and algorithmic penalties for non-disclosure.

The updated policy makes the “AI-generated profile” label effectively mandatory for AI influencer operators seeking growth on Instagram. Accounts that comply with labeling can continue to benefit from algorithmic recommendations and remain eligible for monetization programs, while unlabeled AI persona accounts face systematic demotion, limiting brand discovery and campaign opportunities, according to Meta’s enforcement guidelines. Brand marketers also face tighter compliance obligations, as undisclosed AI in influencer campaigns or ads can trigger demotion or rejection, company records show.

Meta’s technical infrastructure relies on C2PA and IPTC metadata standards to automatically identify AI-generated images and apply labels such as “AI Generated” or “AI info” when provenance signals are present. This approach allows the platform to close loopholes where creators fail to self-disclose AI content, company technical briefs explain. Meta’s policies distinguish between AI personas and AI-altered depictions of real individuals; the latter can violate rules on manipulated media and deepfakes, resulting in stronger enforcement such as content removal rather than just demotion.

The platform’s combination of automatic detection, mandatory labels, and algorithmic demotion is part of a broader strategy to manage AI-generated influencers and profiles amid emerging global AI regulations that emphasize synthetic media transparency and deepfake controls, according to Meta officials and regulatory analysts. Meta’s approach reflects ongoing efforts to balance innovation with user protection and authenticity demands in the evolving social media landscape.

Reddit updated its application programming interface (API) rules again in June 2023 amid protests from thousands of subreddit communities across the platform. The changes followed Reddit’s decision to charge third-party developers for API access, citing high infrastructure costs caused by some apps generating millions in revenue, officials said.

The protests erupted after Reddit announced in April 2023 that it would begin charging third-party developers for API access, ending the previous model of broadly free data use. According to Reddit’s official communications, apps requiring higher usage limits would be charged $0.24 per 1,000 API calls, an amount the company framed as “less than $1 per user per month” for a typical third-party Reddit app. The new pricing was set to take effect July 1, 2023. Reddit indicated that apps making fewer than 100 API calls per minute would continue to have free access within published rate limits, with fees primarily targeting high-volume, often monetized applications, officials said.

On June 12, 2023, thousands of subreddits—including some of the platform’s largest communities—went private or restricted posting to protest the new rules.

Reddit emphasized that moderator tools, bots, and certain noncommercial accessibility apps would retain free API access, carving out exemptions for functions considered critical to community operations and accessibility. The company justified the changes by citing infrastructure costs of approximately $10 million annually, which it attributed to some third-party developers generating “millions” in revenue from their apps, according to statements from Reddit CEO Steve Huffman and company officials.

The API pricing announcement triggered widespread backlash from Reddit’s volunteer moderators and users. Reports from multiple sources cited figures ranging from 7,600 to over 8,000 subreddits participating in the blackout. Some communities planned at least a 48-hour blackout, while others extended their private status past Reddit’s proposed deadline, allowing the protest to continue beyond the initial window. By late June, estimates indicated nearly 9,000 subreddits had gone dark during the protests, underscoring the breadth of opposition, according to Reddit community records and media reports.

Moderators and protest organizers argued that the pricing would effectively shut down popular third-party apps used for moderation, browsing, and accessibility, thereby harming users who rely on these tools. Developers warned that the fees—amounting to $12,000 per 50 million API requests—would translate into tens of millions of dollars annually for large consumer apps like Apollo. Media coverage highlighted concerns that the new costs would “effectively kill off” many third-party applications that offer alternative interfaces and advanced features.

In response to the backlash, Reddit reiterated that the vast majority of API users would not have to pay for access, emphasizing that noncommercial, accessibility-focused apps could continue to access the API for free. Reddit officials confirmed that agreements had been reached with some accessibility-focused apps and essential tools, including Pushshift, to maintain access. The company also announced plans to roll out new first-party accessibility and moderation tools in stages through August 2023, including updates to moderation controls, messaging, and user approvals and bans.

The ongoing blackout led Reddit to press moderators to reopen subreddits, warning that prolonged private status could conflict with Reddit’s requirement that communities remain available to users. Reports indicated that Reddit staff warned some communities they might replace or remove moderators if subreddits remained dark indefinitely. This enforcement stance highlighted tensions between Reddit’s corporate governance and volunteer moderators, who argued that blackouts were necessary leverage to protest changes they saw as harmful.

The dispute over API pricing and access also drew attention to Reddit’s reliance on volunteer moderators who depend heavily on third-party tools and data access to manage large, complex subreddits. Analysts and community members noted that the conflict was not only about pricing but also about control over data, tools, and labor, with moderators asserting their role in maintaining the platform’s value.

Looking ahead, subsequent reporting in 2026 described a “Responsible Builder” policy requiring every Reddit API app owner to register by September 30, 2026, to remain in good standing and be included in Reddit’s future roadmap. Technical analyses showed that the public Data API remained operational for existing approved apps but was closing to new ones, signaling a gradual tightening of access. Reddit announced on August 5, 2026, that it would “gradually start restricting all new requests” to the public Data API and encourage third-party apps to migrate to its Developer Platform, a system for running JavaScript and TypeScript inside Reddit. The company stated that Python support was not on the roadmap, potentially affecting current tooling ecosystems.

Reddit also introduced a $1 million migration program to incentivize selected apps to move to the Developer Platform by December 31, 2026, while clarifying that no fixed shutdown date had been set and existing approved OAuth credentials would continue working “for now,” according to technical reports.

These API and data access changes occurred amid increasing use of Reddit content by AI companies and other large-scale data consumers, who now face new costs to access Reddit’s “vast troves of data.” Starting July 1, 2023, third-party developers requiring higher usage limits—including firms like OpenAI—must pay the new fees, according to Reddit’s official statements. Commentators linked the API pricing controversy to a broader debate over monetizing user-generated content heavily used for training AI models and other data-intensive applications.

The Reddit API dispute has drawn attention as a case study in how social platforms balance revenue, infrastructure costs, third-party innovation, and community governance in an era of expansive data use. The evolving policies and enforcement actions illustrate ongoing challenges in managing data access and platform control amid competing interests from corporate operators, volunteer moderators, developers, and users.

Mastodon instance administrators coordinated new shared blocklists in early 2023 to address harassment targeting activists on the platform. According to Mastodon’s official documentation, the blocklists allow admins to suspend or limit problematic servers in bulk, streamlining moderation efforts against hate speech and harassment.

According to Mastodon’s official documentation, the feature allows admins to upload CSV files containing lists of other servers to suspend or limit in bulk, streamlining moderation efforts against hate speech and harassment. The documentation outlines a standardized process where admins download a blocklist, navigate to Preferences → Moderation → Federation, import the list, and review entries before applying blocks. Admins can also export their own blocklists to share with others.

Starting in early 2023, Mastodon introduced an instance-level blocklist import and export feature designed to help administrators manage problematic servers more efficiently.

This feature was developed in response to the challenge of handling “hundreds of problematic instances,” which made manual blocking tedious and prone to error, especially for new administrators unfamiliar with which servers to block, according to Mastodon developers and community researchers. Shared blocklists maintained by individuals such as Rohan Kumar (known as Seirdy), Gardenfence, and Oliphant have gained significant adoption since the feature’s rollout, according to a 2023 research overview of fediverse moderation practices.

Several public blocklists have emerged with a focus on protecting marginalized users from harassment and hate speech. The Gardenfence blocklist, hosted as a GitHub project, describes itself as “a basic fediverse blocklist for Mastodon admins to block the most well known sources of hate speech and harassment,” targeting instances associated with racism, sexism, homophobia, transphobia, doxxing, stalking, and trolling. Another list, Fblocklist, launched in January 2024, similarly aims to block servers known for hate speech and harassment. The JoinFediverse wiki documents multiple public blocklists used by admins, including FediSeer by Rikudou, Gardenfence by Brook, Oliphant’s blocklists, Seirdy’s blocklist, and The Bad Space by Ro.

The FediBlock ecosystem exemplifies community coordination among instance administrators and users sharing information about abusive or malicious servers. Research shows that admins aggregate multiple sources—including stux’s list from mastodon.social, hashtag-driven reports like #fediblock, and older curated lists—to create composite blocklists published on platforms such as GitHub and Codeberg. Mastodon GitHub records dating back to August 2019 reveal early interest in sharing mutelists and blocklists between “friend instances” to facilitate coordinated moderation.

Harassment controversies targeting activists and marginalized groups have been a significant driver behind the adoption of shared blocklists. Research on fediverse moderation highlights campaigns involving the #asstodon hashtag, where activists became targets of coordinated harassment on Mastodon and other ActivityPub platforms. Commentators have warned that the fediverse can be “unsafe by design and unsafe by default” for marginalized communities, citing the presence of hundreds of problematic instances. Blocklists like Gardenfence explicitly state their goal of protecting “non-white, non-straight, non-binary folk” from servers known to cause drama and reduce safety on the platform. Oliphant’s blocklists categorize instances with reasons such as “Harassment” and “Hate speech,” suspending servers associated with stalking, doxxing, and attacks on marginalized users.

Technically, Mastodon’s blocklist feature uses a CSV format in which each row lists an instance domain alongside a moderation action—such as “suspend” or “silence”—and a reason, including categories like harassment, hate speech, or spam. For example, Oliphant’s blocklist includes multiple domains related to well-known harassment hubs such as kiwifarms.cc, marked for suspension due to harassment. The lists also classify entries under categories like “Third-party bots,” “Mirroring accounts without permission,” and “Spam,” allowing admins to review and select blocks based on detailed reasons. Research indicates that while many moderators share their blocklists publicly as references, others maintain private lists, reflecting a range of coordination styles.

Despite their widespread use, shared blocklists have raised governance concerns within the Mastodon community. Analysts caution that widely shared blocklists can introduce harms if they reflect biased judgments or are applied without local review, potentially affecting already marginalized communities. Academic research notes that a small number of list maintainers can exercise considerable influence over which communities are defederated, raising questions about delegated moderation and power asymmetries. Proposals for themed blocklists—such as those focused on racism—suggest a model where instances subscribe to curated lists to block servers for specific harms, but critics warn this may risk mislabeling or over-blocking activist spaces, some of which are run by marginalized groups.

Historically, early efforts to curate safer spaces on the fediverse began around 2019 with projects like FediHealth, which suggested instances to follow or block based on federation policies. Calls for shared blocklists intensified from 2022 onward as administrators sought collective tools to manage racist and harmful content more effectively. By 2023, blocklists maintained by Seirdy, Gardenfence, and Oliphant became key resources for defederating from poorly moderated or racist servers. In 2024, additional lists such as IFTAS’ Do Not Interact (DNI) and CARIAD were introduced, often emphasizing protection for marginalized and activist communities.

Overall, Mastodon instance administrators continue to coordinate through shared blocklists to respond rapidly to harassment campaigns targeting activists and marginalized users. This coordination occurs alongside ongoing debates about governance, fairness, and the balance between protecting users and preserving open federation.

Meta Platforms agreed Monday to pay up to $18 billion to settle lawsuits brought by a bipartisan coalition of U.S. states over claims that Facebook and Instagram were designed to addict children and harm their mental health. According to officials, the settlement resolves a multistate lawsuit and includes guaranteed payments over 10 years, avoiding a landmark trial in California federal court.

Under the terms of the settlement, Meta will pay the states up to $18 billion over a 10-year period, with $12.7 billion guaranteed regardless of other platforms’ actions, and about $5 billion contingent on similar commitments from competitors such as YouTube, TikTok, and Snap, according to officials familiar with the agreement. The funds are designated for youth online safety programs and related state priorities, rather than direct compensation to individual users, Meta said in a statement released Monday.

The settlement resolves a multistate lawsuit led by a bipartisan coalition of state attorneys general, involving between 48 and 52 states plus the District of Columbia and several territories, depending on the source.

The original 2023 suit was filed by 29 states alleging that Meta designed Facebook and Instagram to be addictive for children and teens, contributing to mental health harms including anxiety, depression, and body-image issues. Additional states and territories later joined, creating a near-nationwide settlement, according to court records.

The agreement ends a landmark trial scheduled in California federal court, which would have featured testimony from senior Meta executives including CEO Mark Zuckerberg, sources confirmed. Meta resolved related claims across multiple state groupings under a single framework, capping its total financial exposure at $18 billion over the decade.

As part of the settlement, Meta will implement specific safety measures for users aged 13 to 17 on Facebook and Instagram. These include a default two-hour daily time limit on app use, which only parents can modify or disable, officials said. Teen accounts will also receive prompts after 15 minutes of continuous use encouraging breaks, and a “school mode” will reduce or mute push notifications during typical school hours, such as 8 a.m. to 3 p.m.

The agreement mandates “night mode” or overnight blocks that prevent teen access by default between midnight and 6 a.m., with the possibility of broader restrictions if other platforms adopt similar policies. Additionally, like and reaction counts will be hidden by default on posts made by or visible to teen users to reduce social comparison pressures, according to sources familiar with the settlement.

Meta will offer an optional non-personalized feed for teen users that does not rely on engagement algorithms or behavioral profiling, which the company must roll out within four months of the settlement taking effect. The company also agreed to limit recommendation and targeting systems aimed at teens, including disabling features that promote potentially harmful or addictive content, officials said.

Enhanced age assurance measures are a key component of the deal. Meta will deploy technical age-checking systems and policy enforcement to prevent children under 13 from accessing Facebook or Instagram and to restrict teen access to inappropriate content. Major age-assurance requirements must be implemented within one year, with verification and audit provisions allowing state oversight to ensure compliance, according to court documents.

The multistate lawsuits alleged that Meta knowingly designed addictive features such as infinite scroll and algorithmic feeds to increase engagement among minors while disregarding internal research on potential harms. Attorneys general accused the company of violating consumer protection and child safety laws by failing to warn users and by misrepresenting the risks of its platforms to young people. The case attracted national attention as one of the highest-profile legal challenges to social media’s impact on children.

In its statement, Meta denied wrongdoing and did not admit liability but expressed commitment to improving safety for young users. State officials described the settlement as a historic agreement that will reshape how teens interact with social media. The deal also sets a benchmark for other platforms, as part of Meta’s payment is contingent on rivals adopting similar safeguards and making comparable payments to states.

The settlement is structured as a consent judgment, allowing courts and attorneys general to enforce the terms if Meta fails to implement the promised changes. It also consolidates multiple lawsuits into a single framework, providing a coordinated approach to addressing youth online safety across the United States.

The broader implications of the case include establishing new regulatory expectations for social media companies regarding child safety and mental health. The settlement’s enforcement mechanisms and transparency requirements aim to ensure ongoing compliance. The multistate coalition and Meta will oversee implementation, with periodic reporting and audits to monitor progress.

Meta agreed Wednesday to limit daily use of Facebook and Instagram to two hours for users under 18 in the United States as part of a $16 billion settlement with state attorneys general. The restriction, described as a “hard limit” by officials, aims to address allegations that the platforms were intentionally addictive to teens and will require parental involvement to override.

Meta will track usage across multiple accounts linked to the same teen to prevent circumvention by switching profiles, officials said. The restriction is described as a “hard limit,” meaning it is enabled by default and cannot be disabled by the teen without explicit parental permission.

The two-hour daily limit applies to all Facebook and Instagram users under 18 within the United States, combining time spent across both platforms, according to court documents and statements from state attorneys general involved in the settlement.

In addition to the daily cap, the settlement requires Meta to implement a default overnight block from midnight to 6 a.m. for teen users, effectively restricting access to most app features during those hours. Messaging functions, however, are exempt from both the daily limit and overnight curfew, allowing teens to send and receive direct messages at any time, court filings show. The overnight block can only be adjusted with parental consent, according to state officials.

The settlement also includes a “School Mode” that silences most notifications during typical school hours, generally from 8 a.m. to 3 p.m. on weekdays, to reduce distractions, sources confirmed. During school hours, teens will have limited access to Facebook and Instagram features, though basic functionality remains available within the daily time allowance. These notification controls form part of a broader suite of teen-safety features, which also include improvements in age verification and hiding “like” counts to mitigate social-comparison pressures, according to Meta and state releases.

Meta has agreed to roll out these changes within one year of the settlement’s approval, with some provisions expected to take effect within six months in participating states and territories. The settlement, announced Wednesday, August 26, 2026, involves coordination with roughly 29 to 50 U.S. states and territories and awaits approval by a federal judge in Oakland, California. Once approved, the restrictions and behavioral changes will remain in place for between five and ten years, depending on the specific provision, officials said.

The agreement is valued at approximately $16 billion over a decade, combining direct payments and mandated platform changes designed to address allegations that Meta’s products were intentionally addictive to teens. Meta has denied wrongdoing but committed to the terms as part of the settlement, according to company statements and court records.

The settlement includes conditional provisions that would further tighten restrictions if rival platforms such as TikTok, YouTube, and Snapchat adopt similar time-limit commitments. Under those circumstances, Meta would reduce the teen daily limit from two hours to one hour per day per app, with a maximum of two hours total across all participating platforms. The overnight block would also extend from 10 p.m. to 7 a.m., lengthening the curfew window. These conditions aim to encourage industry-wide standards, according to state attorneys general.

To enforce the time limits, Meta will introduce forced pauses and prompts within the apps, including alerts every 15 minutes of continuous use and additional warnings at 60 and 90 minutes. Once the two-hour limit is reached, teens will be blocked from further use of the main app experiences until the next day unless a parent authorizes additional time, court documents show. The daily allowance resets at midnight.

The settlement also mandates strengthened age estimation mechanisms to prevent children under 13 from accessing Meta’s platforms and to ensure teens are assigned to age-appropriate accounts before time limits and other safeguards apply. Compliance with the settlement will be monitored by state attorneys general and the court, which retain authority to review Meta’s adherence and require ongoing enforcement of the teen-safety measures for the agreed duration.

Certain exceptions to the time limit include direct messaging and long-form content, such as extended videos, which do not count toward the daily cap, according to court filings and official statements. The specifics of these exemptions are detailed in state attorney general releases and court documents but are not fully public.

The settlement’s implementation timeline anticipates that the new restrictions and features will be operational within the next year in participating jurisdictions, with some elements potentially launching within six months. The court’s approval is the final step before these measures become enforceable.

Twitch experienced a major outage affecting its web, chat, and video-watching services on Thursday, August 27, 2026, shortly after the GTA 6 “Extended Look” gameplay trailer premiered on Netflix. According to Twitch’s status page, the disruption began around 3:15 p.m. ET and was under investigation as users reported login difficulties and impaired chat functionality.

The outage began shortly after 3 p.m. Eastern Time, coinciding with the premiere of the GTA 6 “Extended Look” gameplay trailer on Netflix. Reports of Twitch service disruptions surged on Downdetector between 3:15 and 3:30 p.m. ET, with Twitch acknowledging the issues on its status page shortly after 4 p.m. ET. The platform’s status page classified the incident as a “major outage” affecting core services including the website, chat, and video watching, while video broadcasting capabilities remained operational, allowing streamers to continue transmitting content despite widespread viewer difficulties.

At its peak, Downdetector recorded more than 17,000 user complaints related to Twitch, citing problems such as failed logins, black screens on streams, error messages, and impaired chat functionality.

Users reported intermittent access to streamer profiles and broken search features, with some navigation pages like Browse remaining partially accessible. Regional media in Chile reported connectivity issues between 3 p.m. and 4 p.m. local time, consistent with the global timeline of disruptions. Twitch’s status page updates indicated that while many services began to recover within 30 minutes to an hour after the worst of the outage, residual problems persisted into the early evening.

Twitch officials confirmed the outage on social media and its status page, stating, “We are aware of issues impacting several areas of our site. Our teams are working to resolve these.” The company did not provide a specific cause for the disruption and did not attribute the outage directly to the GTA 6 trailer release. Twitch also announced that the GTA VI game category would remain enabled for creator streams until 9 p.m. Pacific Time on the day of the extended look, signaling ongoing support for related content despite the technical difficulties.

Several prominent creators, including IShowSpeed and Kai Cenat, streamed live reactions to the GTA 6 trailer on Twitch at its 3 p.m. ET release. However, by approximately 4:45 p.m. ET, their livestreams and profiles became inaccessible, according to reports from Forbes and other outlets. Eurogamer and PC Enthusiast noted that Twitch went down during one of the largest GTA 6 streaming events to date, though Twitch had not confirmed any direct link between the trailer’s popularity and the outage.

Media coverage from Forbes, Newsweek, Eurogamer, and GamesRadar consistently framed the outage as occurring soon after or during the GTA 6 extended trailer event. Forbes highlighted the spike in user complaints, while GamesRadar contrasted Twitch’s failure with Netflix’s stable handling of the trailer premiere. ADN Radio in Chile described the event as “provocó la caída” (caused the fall) of Twitch, reporting intermittent issues coinciding with the local time of the trailer’s launch.

Despite speculation, Twitch has not publicly disclosed the technical cause of the outage. Several outlets, including Mashable and PC Enthusiast, emphasized that while the timing aligned with the GTA 6 trailer release, there is no confirmation that high traffic or GTA-related content directly caused the disruption. Minor and brief outage spikes were noted for Netflix at the trailer’s launch on Downdetector, but these did not escalate into a sustained platform-wide incident.

Twitch’s diagnostics showed that while web, chat, and video watching services were severely impacted, video broadcasting remained functional. Users experienced login difficulties, impaired search capabilities, and unstable or nonfunctional chat systems. Some site areas, such as the Browse page and category listings, remained accessible even as individual streams displayed black screens or failed to load. The Twitch status page maintained outage flags beyond the initial recovery period, indicating ongoing or residual issues with certain subsystems.

The outage occurred during a highly anticipated gaming event, drawing significant attention from Twitch’s global user base. Twitch’s teams continued investigating the disruption as of the evening of August 27, 2026, with no further details released regarding the root cause. The platform’s support for GTA VI streaming content remained in place through the night, reflecting efforts to maintain engagement despite technical challenges.

On August 25, 2026, Google confirmed that YouTube expanded its Custom Feeds experiment to support multiple custom feeds per account for signed-in users already in the trial. According to the company, this update builds on the initial single-feed feature introduced in May 2026 and is available on both mobile and desktop as part of an ongoing experiment with no general release date announced.

These feeds appear as labeled chips aligned horizontally at the top of the YouTube Home screen on both mobile and desktop platforms, enabling quick toggling between multiple personalized recommendation streams, according to Google’s August 25, 2026, announcement. Users can generate each feed by entering a text prompt describing the desired video content or by selecting from suggested prompts, with YouTube’s AI then curating a dynamic recommendation list based on the prompt, watch history, and search activity.

The update allows users in the experiment to create and maintain up to eight separate custom feeds, each tailored to different topics or moods.

This expansion reverses the earlier limitation, introduced in May 2026, which restricted accounts to a single active custom feed at a time. Previously, users had to overwrite the existing feed’s prompt to change topics, but the new multi-feed capability lets participants maintain several concurrent feeds, each with its own unique prompt and recommendations, sources confirmed. The company noted that inactive feeds expire after a few weeks of non-use to prevent clutter and encourage users to keep their feeds relevant, with the expiration timeframe reported by multiple outlets as roughly 30 days.

Access to multiple custom feeds remains limited to signed-in users in the United States who were already part of YouTube’s original Custom Feeds experiment. The test group must have search and watch history enabled, as these data points are essential for the AI-driven recommendation system to function effectively, according to official YouTube help documentation and statements from Google representatives. The feature is currently available only in English and is described by the company as an ongoing experiment without a set date for general release.

The interface changes include a horizontal row of custom feed tabs resembling topic-based timelines found on other social platforms. Selecting a chip loads a full-screen feed of videos tailored to the prompt, visually similar to the standard Home feed but scoped to the user’s chosen theme. Each custom feed includes a text box at the top where users can edit the prompt to refine or change recommendations instantly without deleting the feed. This functionality allows for highly specific prompts, such as “competitive Valorant guides under 15 minutes” or “guided meditations under 10 minutes,” producing feeds aligned with detailed viewer intent, according to coverage by tech news outlets.

Earlier coverage of the single-feed experiment, which began in November 2025 and rolled out broadly in May 2026, emphasized that users could pin their custom feed beside the regular Home feed but could not maintain multiple feeds simultaneously. The new multi-feed update addresses this by allowing multiple “mood- or topic-based feeds” to coexist, a capability frequently requested by users, sources said. The system’s design ensures that feeds tied to signed-in accounts can influence overall recommendation patterns through watch history and prompt edits, though each feed remains scoped to its specific prompt during use.

YouTube officials have framed the multi-feed experiment as part of a broader AI-powered personalization initiative, allowing users to steer recommendation algorithms with natural-language prompts rather than relying solely on passive viewing behavior. Analysts and commentators have compared the feature to “Bluesky-style custom feeds,” noting that it offers more granular control over content consumption by organizing recommendations into separate, labeled feeds. This approach is seen as a competitive response to other platforms experimenting with user-curated or algorithmically distinct feeds, aiming to retain viewers within YouTube’s ecosystem across different interests.

The company has not provided a firm timeline for when multiple custom feeds will be made available to all users globally. Instead, it continues to describe the feature as an experimental “playground” with potential changes before any wider release. The current rollout remains confined to a limited test group, primarily in the United States, with English-language support only. Users outside the experiment or those who do not meet the eligibility criteria cannot access the multi-feed feature at this time, according to Google’s public statements and help resources.

The original Custom Feeds feature, introduced broadly in May 2026, marked YouTube’s first step toward user-driven, AI-powered feed personalization. The expansion to multiple custom feeds announced in August 2026 builds on that foundation, reflecting ongoing efforts to enhance user control over video recommendations. The system’s reliance on watch and search history underscores the importance of user activity data in shaping personalized content, while the automatic expiration of inactive prompts aims to maintain an organized and relevant user interface.

Take-Two Interactive, parent company of Rockstar Games, filed subpoenas on August 20, 2026, in the U.S. District Court for the Southern District of New York seeking information from Microsoft and Discord. According to court filings, the subpoenas aim to identify individuals responsible for leaking copyrighted Grand Theft Auto 6 gameplay footage.

The subpoenas, filed on August 20, 2026, seek to identify the individuals responsible for distributing unauthorized Grand Theft Auto 6 gameplay footage, according to court documents reviewed by multiple media outlets. Take-Two Interactive, the parent company of Rockstar Games, requested the subpoenas in the U.S. District Court for the Southern District of New York, targeting Microsoft and Discord as part of an effort to trace the source of the leaks, sources confirmed.

The subpoenas also reference Discord servers such as one run by the user DarkViperAU and another called “!Odyssey,” with Take-Two requesting data on all users who were members of or communicated within these servers, according to coverage from Mashable and PC Gamer.

The legal filings are described as Digital Millennium Copyright Act (DMCA) subpoenas aimed at uncovering the identities of alleged infringers who circulated copyrighted material. Take-Two’s declaration states that the subpoenas seek records connected to accounts and servers believed to be involved in the distribution of the leaked content. The company is requesting information that could link online aliases to real-world identities, including account registration data and communication logs, according to reports from Variety and Game Developer.

Discord was specifically named as a source of user account information in the subpoenas. Reported filings seek identifying details for Discord users associated with the leak, including account names, server participation, and membership in certain channels. Among the usernames mentioned in court filings and media reports are “CYBERLEEK,” “CINEMATICROCKSTAR,” and “Surfer24k.”

The subpoenas demand a wide range of personal and technical information. Reported requests include account IDs, registration emails, IP addresses, phone numbers, and linked accounts such as Google or Xbox profiles. Take-Two is also seeking device identifiers and, in the case of Microsoft, internal investigative records related to the leaked material. The Microsoft subpoena reportedly references files stored on GitHub and OneDrive, according to Engadget and Tom’s Hardware.

The leaked content at the center of the legal action consists of unauthorized gameplay footage from the unreleased Grand Theft Auto VI. Reports indicate that the copyrighted material includes audiovisual content, artwork, images, dialogue, and other creative elements from the game. The footage has been circulating widely on Discord and X, formerly known as Twitter, with the account “CYBERLEEK” linked to posting the unauthorized clips, according to Forbes and GameSpot.

Take-Two has set a deadline for Microsoft and Discord to comply with the subpoenas by September 4, 2026. This timeline reflects the company’s urgent efforts to halt further dissemination of the leaked content. Multiple outlets, including Kotaku and Variety, noted that the roughly two-week compliance window underscores the escalating response to recent GTA 6 leak activity. Neither Microsoft nor Discord has publicly commented on the subpoenas as of the latest reports.

The court filings emphasize that the legal action is grounded in copyright enforcement rather than criminal prosecution. Take-Two is pursuing civil remedies to protect its intellectual property and prevent further unauthorized distribution of its confidential game materials. The subpoenas represent a broader strategy by the company to identify and hold accountable those responsible for the leaks.

The Grand Theft Auto series is a major franchise for Rockstar Games and Take-Two Interactive, with GTA 6 highly anticipated by the gaming community. The recent leaks have prompted swift legal responses aimed at protecting the company’s intellectual property ahead of the game’s official release. The ongoing investigation and court proceedings will determine the extent of the leak’s origins and the identities of those involved.

The U.S. Department of Justice announced a $400 million settlement with TikTok, ByteDance, and affiliated entities on Tuesday to resolve allegations of violating the Children’s Online Privacy Protection Act. According to DOJ officials, the settlement includes $300 million paid immediately and $100 million contingent on vacating a 2019 consent decree against TikTok’s predecessor, Musical.ly, related to children’s privacy violations.

The 2019 decree was the result of an earlier Federal Trade Commission enforcement action addressing violations of the Children’s Online Privacy Protection Act, or COPPA. According to DOJ officials, the settlement resolves allegations that TikTok, ByteDance, and affiliated entities violated COPPA and infringed upon the existing 2019 FTC consent order related to children’s privacy protections.

Under the terms of the settlement announced by the U.S. Department of Justice, TikTok will pay $300 million immediately, with an additional $100 million contingent on a federal court vacating a 2019 consent decree against Musical.ly, TikTok’s predecessor, officials said.

The 2019 consent decree, records show, arose from a case in which Musical.ly agreed in February 2019 to pay $5.7 million—the largest COPPA civil penalty at the time—for collecting personal information from children under 13 without parental consent. The decree required Musical.ly to destroy personal data from users under 13, remove accounts without verifiable age information by May 2019, and implement compliance monitoring and recordkeeping measures. The FTC’s 2019 announcement also noted that the app had to take offline all videos created by children under 13. The new settlement’s $100 million payment is expressly conditioned on a court order vacating this permanent injunction, effectively ending those prior obligations, according to DOJ statements and legal filings.

The vacatur mechanism is a key feature of the settlement, making the final payment dependent on the formal termination of the 2019 decree. DOJ officials explained that this arrangement converts the prior regulatory constraints into a historical matter once vacated, while simultaneously resolving the 2024 enforcement litigation. Legal experts and former FTC officials cited in media reports have noted that vacating the 2019 order would eliminate ongoing compliance reporting, monitoring, and recordkeeping requirements imposed on TikTok’s predecessor.

The 2024 lawsuit, filed in August by the DOJ on behalf of the FTC, alleged that TikTok and ByteDance flagrantly violated COPPA by allowing millions of children under 13 to use the platform and by collecting their personal data without obtaining verifiable parental consent, despite the prior enforcement action. The complaint further asserted that TikTok stored, collected, and processed data from child users in violation of the 1998 federal law. DOJ and FTC officials emphasized that the 2024 case involved alleged repeated violations of the 2019 order, describing the conduct as a massive-scale invasion of children’s privacy through extensive data collection practices.

DOJ’s press release highlighted the settlement as a major enforcement action in the children’s privacy domain, describing it as one of the largest recoveries ever obtained in a COPPA-related case. The department underscored that the claims resolved by the settlement are allegations only, with no determination of liability. News reports characterized the deal as a DOJ-TikTok agreement resolving the 2024 children’s privacy suit and positioning the $400 million payment as a record-scale resolution in the COPPA enforcement landscape.

COPPA, enacted in 1998, requires online services to obtain verifiable parental consent before collecting personal information from children under 13. The 2019 Musical.ly case was the largest COPPA settlement at that time, with a $5.7 million penalty and significant behavioral remedies. Despite that earlier enforcement, regulators alleged in 2024 that TikTok continued to permit underage users and collect extensive data, prompting renewed litigation. The FTC’s 2024 complaint referenced the 2019 permanent injunction (United States v. Musical.ly, No. 2:19-cv-01439-ODW-RAO) and detailed obligations that the new case claimed were not adequately honored.

In addition to the financial terms, DOJ materials indicated that the settlement includes measures to strengthen protections for young users, though public reporting has focused primarily on the monetary components and the vacatur condition. Analysts and officials noted that the settlement reflects broader efforts by federal agencies to enforce COPPA in the social media sector, especially concerning large platforms’ handling of minors’ data.

The settlement concludes the current litigation but leaves open questions about TikTok’s ongoing compliance framework, given the vacatur of the 2019 decree would remove certain regulatory obligations. DOJ and FTC officials have framed the enforcement actions as part of an intensified regulatory focus on children’s online privacy rights, consistent with evolving federal priorities in this area.

A proposed class action lawsuit was filed Thursday in the U.S. District Court for the Northern District of California against Twitch Interactive Inc. and Amazon.com Inc. by Twitch streamer Warren Pandiscia and others. The complaint alleges the companies unlawfully used streamers’ videos without permission to train Amazon’s generative AI systems, violating implied contracts and state unfair competition laws, according to court documents.

The lawsuit, filed Thursday in the U.S. District Court for the Northern District of California, alleges that Twitch and Amazon “unlawfully copied millions of videos” from creators’ streams to develop Amazon’s generative AI models without negotiating licenses or obtaining consent, according to court documents. Lead plaintiff Warren Pandiscia, a Connecticut-based Twitch creator with more than 900 followers, represents a proposed class of Twitch streamers whose content was allegedly used in the AI training process.

The complaint claims that Twitch and Amazon harvested streamers’ video content—including live streams, archived videos on demand (VODs), clips, chat logs, and channel images—starting as early as 2024 during an internal prototyping phase.

Plaintiffs say this data collection occurred without permission or compensation and continued until Twitch introduced an opt-out setting in mid-August 2026. Twitch publicly announced the “Training for Generative AI” setting around August 12–13, 2026, which is enabled by default, allowing Amazon to use eligible channel content for AI training unless streamers actively disable the feature, according to Twitch’s documentation and reporting by Courthouse News and AI Rights Brief.

The complaint argues that because AI models cannot be “un-trained,” creators cannot recover intellectual property already copied and used, even if they opt out later. Plaintiffs assert that Twitch and Amazon’s conduct violates implied contracts with streamers and state unfair competition laws, depriving creators of control over their works, economic value, and property rights. The lawsuit seeks injunctive relief to halt the use of streamers’ content for AI training without authorization, as well as damages, restitution, and disgorgement of profits allegedly earned from the AI models.

According to the complaint, Amazon sought AI training data “at an unprecedented rate” and therefore relied on Twitch streams instead of obtaining lawful licenses or creator consent. The lawsuit describes the defendants’ AI data pipeline as enrolling “every creator” by default through a setting activated on August 12, 2026. Twitch has stated it does not directly train AI models but permits Amazon to use Twitch content for generative AI training under the default-on setting. The opt-out process requires streamers to navigate to Settings → Security and Privacy and toggle off the “Training for Generative AI” option, which is located near the bottom of the menu.

The filing also highlights that Twitch confirmed in August 2026 that it uses creators’ content for AI training and updated its Terms of Service and Privacy Policy to reflect this practice. Forbes reported that Twitch internally acknowledged “nobody would opt in” to AI training, which influenced the decision to make the setting default-on and require opt-outs rather than opt-ins. Critics have noted that the setting is buried in privacy menus, making it unlikely many streamers would find and disable it without explicit guidance.

The lawsuit follows significant backlash among Twitch streamers and viewers after the opt-out policy was announced. Media outlets including BBC and NPR described the reaction as “furious streamer backlash,” with creators criticizing the default-on approach and lack of prior notice. The complaint references Twitch’s August confirmation and the ensuing controversy as part of the context for the legal action.

The case, captioned Pandiscia v. Twitch Interactive, Inc. and Amazon.com, Inc., remains at an early stage as a proposed class action. Class certification has not yet been granted. The Northern District of California is a common venue for technology-related class actions, and the complaint focuses specifically on Amazon’s use of Twitch streamers’ content to train generative AI, rather than broader platform issues.

Plaintiffs allege that Twitch and Amazon engaged in deceptive and unfair business practices in violation of state consumer protection laws. The complaint states that the defendants’ systems “are incapable of obtaining the consent of all parties to the communications they capture,” referring to streams and associated content. The lawsuit seeks to hold Twitch and Amazon accountable for what it describes as unauthorized use of creators’ intellectual property to develop AI technology.

The outcome of the case could have implications for how online content platforms handle user-generated material in the context of AI training. The lawsuit adds to ongoing legal and public debates over data use, consent, and compensation in the rapidly evolving field of generative artificial intelligence.