Facebook is a social media and social networking service owned by Meta Platforms. Launched on February 4, 2004, by Mark Zuckerberg and fellow students at Harvard University, it developed from a university networking website into an international platform. Its services encompass personal profiles, shared posts and audiovisual content, interest-based communities, business pages, and local commerce. Facebook remains the name of the service; its operating company, formerly Facebook, Inc., adopted the name Meta Platforms, Inc. on October 28, 2021. (meta.com)
Facebook combines connections among people and organizations with algorithmically selected content and advertising. Its scale makes it an important subject in the study of online communication, digital advertising, and data privacy: Meta reported 3.07 billion Facebook monthly active users as of December 31, 2023, under its reporting definition. Understanding the service requires distinguishing its products and technical systems from the wider corporate group, and examining its development, audience, commercial organization, content governance, and documented regulatory history. (about.fb.com)
Origins and historical development
Facebook’s name derives from the “face books” used by educational institutions: directories pairing students’ photographs with their names. On February 4, 2004, Mark Zuckerberg launched TheFacebook at Harvard University, initially restricting membership to Harvard students. The website combined directory information with profiles and connections between users; early functions included finding classmates, adding friends, and sending “pokes.” Zuckerberg wrote the initial software. His founding student collaborators contributed different kinds of work: Eduardo Saverin supplied early financing and handled business and advertising matters; Dustin Moskovitz helped program the service and extend it to other universities; Chris Hughes worked on communications and became its spokesperson; and Andrew McCollum contributed the original logo and graphics. These contributions developed during the early operation of the site rather than constituting identical roles at its launch. (en.wikipedia.org)
Expansion initially proceeded through separate university networks. In March 2004, TheFacebook extended access to Columbia, Stanford, and Yale, followed by additional colleges. Zuckerberg and several collaborators moved to Palo Alto, California, that summer, establishing an operating base in Silicon Valley. In 2005, the service adopted the shorter name Facebook and introduced access for high-school students. The decisive removal of its educational membership boundary came on September 26, 2006, when registration opened to anyone aged at least 13 with a valid email address. Membership no longer depended on attendance at a participating institution; regional networks also provided a way to organize connections outside schools and workplaces. (en.wikipedia.org)
Product changes shifted Facebook from a directory-centered website toward an ongoing stream of shared activity. News Feed, launched on September 5, 2006, assembled updates about friends on the home page, reducing the need to visit their profiles individually. The Like button, introduced on February 9, 2009, added a lightweight way to respond to content without writing a comment. Timeline reorganized personal profiles around a chronological record of posts, photographs, and life events. It was introduced at the f8 developer conference on September 22, 2011, but became available worldwide on December 15: the announcement and general availability were separate milestones. During that rollout, users received a seven-day preview period to review their Timeline before its publication. (meta.com)
Mobile access developed alongside these changes rather than beginning with a single smartphone application. Facebook for Mobile launched in April 2006, while the iPhone application followed on July 10, 2008. Access through applications and mobile-optimized websites subsequently became a major part of service use. Facebook’s 2012 annual report recorded 680 million monthly active users of its mobile products in December 2012, compared with 432 million in December 2011. These figures measured mobile activity, not an additional population separate from people using desktop computers: the same account could access Facebook through both. (meta.com)
Facebook, Inc. became publicly traded on May 18, 2012, with shares trading on Nasdaq under the symbol FB. The offering was priced the previous day at US$38 per share and comprised approximately 421.2 million shares, including 180 million offered by the company and approximately 241.2 million by existing shareholders. This was a corporate financing milestone, not a relaunch of the social network. On October 28, 2021, Zuckerberg announced the corporate brand Meta, and Facebook, Inc. became Meta Platforms, Inc. The change distinguished the company’s broader collection of applications and technologies from the Facebook service, whose name remained unchanged; Instagram and other company products did not thereby become parts of the Facebook social network. (en.wikipedia.org)
Profiles, relationships, and community organization
A personal profile represents an individual within Facebook’s social-media network. It combines a name, profile and cover photographs, biographical information—such as education, employment, hometown, and relationship status—and a timeline of posts and other activity. The timeline organizes material associated with that profile, rather than serving as the personalized stream of material from many sources shown in Feed. Profile information and posts can have different audiences: displaying some information publicly does not necessarily make every post public. Friendship is reciprocal, established when one person accepts another’s request. Following is directional: it enables someone to receive another profile’s public updates without establishing a mutual friendship or gaining access to friends-only material. (en.wikipedia.org)
Pages provide public-facing presences for businesses, institutions, brands, artists, public figures, and other organizations. Unlike a personal profile’s reciprocal friendship network, a Page builds an audience of followers. Its identity represents the organization or public presence rather than the individual managing it. Pages and Groups also organize participation differently: Page managers generally publish material on behalf of the represented entity, whereas Groups support discussions initiated by multiple members. Tufts University’s guidance illustrates this distinction by contrasting institutional Pages with Groups created for incoming undergraduate classes, where future classmates can meet and converse. (supremecourt.gov)
Groups organize communities around shared interests, affiliations, or activities. In August 2019, Facebook replaced the earlier public, closed, and secret classification with public and private settings, separating content privacy from discoverability. A private Group can therefore be visible in search without exposing its discussions to non-members, or hidden from general discovery. Membership and permission to participate are also distinct. Facebook’s documentation allows visitors as well as members to contribute to public Groups, subject to administrator controls; participant approval can require review before someone’s first contribution, while post approval can place subsequent submissions in a review queue. (about.fb.com)
Group administrators control settings and appoint administrators or moderators; moderators assist with managing participation and discussions without receiving the same authority over the Group’s configuration. Community tools extend beyond individual discussion threads. At the October 20, 2022 Facebook Communities Summit, Facebook head Tom Alison announced Reels in Groups and updates to Admin Assist, which automatically applies criteria selected by administrators. A daily digest was intended to make those automated actions easier to review. The announcement separately identified experimental features, including contribution-based recognition and changes to Group profiles; these were tests rather than universally available capabilities at that date. (facebook.com)
Events provide a separate structure for organizing gatherings, distributing invitations, collecting responses, and maintaining event-related discussions. Invitations and notifications direct attention to an event, but its privacy setting determines who can view its contents. Facebook distinguishes private, public, friends-only, and Group events. Private events are visible to invited people; public events can be viewed on or off Facebook, and events hosted by Pages are public. Group events inherit visibility from the Group’s privacy setting. Hosts can also control matters such as guest-list visibility and whether discussion posts require approval. (en.wikipedia.org)
Messenger supplies a conversational channel distinct from publishing on a profile, Page, or Group. Introduced as a separate iPhone and Android application on August 9, 2011, it extended Facebook’s existing messaging functions to one-to-one and small-group mobile conversations. Its launch announcement, written by Lucy Zhang, emphasized shared conversation histories across mobile and web access, group planning, and photo attachments. A reply in a Messenger conversation is addressed to that conversation’s participants rather than automatically becoming a public comment or timeline post; similarly, an event invitation or friendship connection is not itself a private-message exchange. (about.fb.com)
Publishing, media, and content discovery
Facebook supports text updates, photographs, videos and links to material elsewhere on the World Wide Web. Original publication creates a post under the publishing account’s identity; redistribution through the Share function circulates an existing post, potentially with additional commentary, rather than making the sharer its original author. Comments attach discussion to a post, while reactions provide a separate, compact response. On February 24, 2016, Facebook expanded Like with five additional reactions—Love, Haha, Wow, Sad and Angry—available through a long press on mobile or by hovering over Like on desktop. These functions distinguish publishing, discussing, responding to and redistributing content, although a single post can support all four activities. (en.wikipedia.org)
Stories and live video introduced different temporal arrangements for publishing. Launched in the main Facebook application on March 28, 2017, Stories assembled photographs and videos into collections viewable for 24 hours. At introduction, they appeared above News Feed rather than as ordinary Timeline or feed posts, unless the publisher also posted the material there. Facebook Live, introduced on August 5, 2015, initially allowed public figures to broadcast through the Facebook Mentions application; wider access followed in 2016. Live streaming allowed viewers to comment, like and share during transmission, while recordings could remain available afterward. Thus Stories emphasized temporary availability, whereas Live emphasized simultaneous transmission and interaction, without necessarily making the recording temporary. (about.fb.com)
Reels launched in the United States on September 29, 2021, as an edited short-video format incorporating music, original audio, effects and other creative tools. Facebook expanded availability to more than 150 countries on February 22, 2022. Initially discoverable through News Feed and Groups, Reels subsequently gained placements in Stories and Watch, illustrating how a publishing format could circulate across several interface locations. In October 2024, Meta announced an updated full-screen Video tab bringing short, long and live videos into one viewing experience. On June 17, 2025, it announced a gradual global transition under which newly uploaded videos would be published as reels without length or format restrictions, alongside renaming the Video tab Reels. This broadened the label beyond its original short-video meaning. (about.fb.com)
The Home and Feeds arrangement announced on July 21, 2022, separated personalized discovery from access organized around existing connections. Home became the name of the application’s opening tab, combining updates from connections with recommended content and access to Stories and Reels. Its ordering used machine-learning predictions rather than publication time alone. Feeds, by contrast, offered the most recent posts from friends, Pages and Groups, with filters and a Favorites list of selected friends and Pages. At launch, Meta specified that Feeds contained advertisements but excluded “Suggested For You” posts: chronological access to connected sources did not mean an advertising-free interface. (about.fb.com)
Search provides query-driven information retrieval alongside these browsing surfaces. Facebook’s October 2015 search expansion included public posts and posts shared by friends, as well as conversations surrounding shared links. Geographic and interest-based discovery were the subjects of further experiments announced on October 4, 2024. The Local tab combined material from Marketplace, Groups and Events, such as nearby activities and items for sale; its announced test covered ten US cities, including Austin, New York City, Chicago and Phoenix. The separate Explore test personalized content around interests such as furniture restoration or marathon training. These were explicitly announced as tests, not universally available tabs, and differed in organizing principle: Local emphasized nearby opportunities, while Explore emphasized topical interests. (about.fb.com)
Audience, international reach, and measurement
Meta defines a Facebook daily active user (DAU) as a registered, logged-in user who visits Facebook through its website or a mobile device, or uses Messenger while also being a registered Facebook user, on a given day. A monthly active user (MAU) meets the corresponding activity requirement during the preceding 30 days. Thus, qualifying Messenger activity can count without a Facebook visit. Reported DAUs are generally daily averages over a specified month, whereas MAUs describe a rolling 30-day audience at a stated measurement date; neither metric measures time spent. (sec.gov)
These account-based statistics are not a census of distinct individuals. Meta estimated that duplicate accounts represented approximately 10% of worldwide MAUs in the fourth quarter of 2023, and false accounts approximately 4%. These estimates rely on sampled account reviews and judgments about authenticity, rather than complete verification of every account. (sec.gov) Meta’s family-wide measures instead estimate distinct people across Facebook, Instagram, Messenger, and WhatsApp, matching accounts believed to belong to one person. They therefore differ from both Facebook-only counts and a simple addition of the services’ audiences. (sec.gov)
Facebook reported 3.03 billion MAUs on June 30, 2023, placing its monthly audience above three billion. (sec.gov) By December 31, 2023, it reported 3.07 billion MAUs, up 3% year over year; average DAUs during December were 2.11 billion, up 6%. Meta’s family-wide monthly active people figure was 3.98 billion at year-end, not a Facebook audience estimate. (sec.gov)
Geographic distribution was uneven. On December 31, 2023, Facebook reported 1.367 billion MAUs in Asia-Pacific, 1.018 billion in “Rest of World,” 408 million in Europe, and 272 million in the United States and Canada. Its reporting definition included Russia and Turkey in Europe, and Africa, Latin America, and the Middle East in Rest of World. Meta identified India, Bangladesh, and Nigeria as the three largest sources of MAU growth during 2023 and of December DAU growth relative to December 2022. These are absolute audience counts, not population-adjusted adoption rates. (s21.q4cdn.com)
Representative surveys provide a different view of audience demography. Pew Research Center’s report published on November 20, 2025, drew its adoption estimates from 5,022 U.S. adults surveyed between February 5 and June 18, using address-based sampling and weighting to represent the adult population. Overall, 71% said they ever used Facebook. (pewresearch.org) Usage was highest among adults aged 30–49, at 80%, compared with 68% among those aged 18–29, 74% among those aged 50–64, and 57% among those aged 65 or older. (pewresearch.org) A separate survey of 5,123 adults, conducted February 24–March 2, found that 52% used Facebook daily, including 58% of 30–49-year-olds and 54% of 50–64-year-olds. These percentages describe all U.S. adults in the respective groups, not just Facebook users; they are self-reported national estimates rather than global activity counts from platform records. (pewresearch.org)
Advertising, commerce, and the platform economy
Facebook’s advertising-supported business model sells businesses access to audiences rather than charging for each ordinary social interaction. Advertising revenue depends on both the volume of advertisements delivered and their average price: greater user activity creates additional opportunities to display ads, while advertiser demand influences their value. Meta reported $131.948 billion in advertising revenue in 2023, against consolidated revenue of $134.902 billion. These are corporate figures, not Facebook-only revenue; they include advertising on Instagram and other placements. Across Meta’s Family of Apps, ad impressions increased 28% in 2023 while average price per ad declined 9%. Meta attributed the price decline partly to growth in lower-monetizing regions and formats, including Reels. An impression represents an advertisement’s display, not necessarily a click or purchase. (sec.gov)
Audience selection and advertising delivery are separate processes. Advertisers can specify audience criteria, supply customer information for matching, or seek to reach people who previously interacted with their websites or applications. In a 2018 explanation, Facebook illustrated targeting with a bicycle shop seeking women in Atlanta interested in cycling, and described customer-list matching and website-based retargeting through Facebook Pixel. These choices establish potential audiences; they do not mean that every eligible person receives the advertisement. (about.fb.com) Meta’s documented auction process combines the advertiser’s bid, an estimated probability of the desired action, and an assessment of advertisement quality. Machine learning supplies predictions used in this process, so automated optimization helps determine actual delivery within the selected audience rather than merely executing a fixed distribution list. (about.fb.com)
Campaign measurement connects advertising expenditure to reported outcomes. Facebook’s reporting distinguishes impressions from engagement measures such as clicks and likes; its broader measurement tools also incorporate information about activity on advertisers’ websites and applications. These measures answer different questions: displaying an advertisement, receiving a response, and recording a subsequent commercial action are not interchangeable outcomes. In its 2019 submission to Australia’s competition regulator, Facebook described reporting metrics based on its own engagement data and aggregated audience insights concerning website visitors, application users, and campaign participants. Advertisers could use these results to change or terminate campaigns. (about.fb.com) This creates a commercial dependency on measurement infrastructure as well as audience access: Meta’s 2023 filing identified restrictions on targeting and measurement signals, including mobile-operating-system changes beginning in 2021, as factors adversely affecting advertising revenue. (sec.gov)
Marketplace added a distinct commerce function. Announced on October 3, 2016, in a launch statement by product-management director Mary Ku, it initially rolled out to adults using Facebook’s mobile applications in the United States, United Kingdom, Australia, and New Zealand. Sellers supplied photographs, descriptions, prices, locations, and categories; buyers browsed nearby listings or filtered searches by location, category, and price. Direct messaging enabled offers and arrangements between buyers and sellers. The original service did not facilitate payment or delivery: its listings connected participants, but a listing or conversation was not equivalent to a transaction completed through Facebook checkout. (about.fb.com)
Marketplace also illustrates how network effects and established distribution can shape competition. In its November 14, 2024 administrative decision, the European Commission found that tying Marketplace to Facebook gave the classified-advertising service a distribution advantage competitors could not match, because Facebook users automatically received access and regular exposure to it. The Commission also found that Meta imposed unfair conditions on competing classified-advertising providers advertising on Facebook and Instagram, allowing their advertising-related data to benefit Marketplace. It imposed a €797.72 million fine. These findings concerned specific uses of market power—including relationships in which competing services were also advertising customers—not a general prohibition on operating an integrated marketplace. (ec.europa.eu)
Recommendation systems and technical infrastructure
Facebook’s social graph represents people, content, actions, and relationships as nodes and edges. A person, photograph, or comment can be an object, while friendship, authorship, and liking can be represented by connections between objects. This stored structure is distinct from a recommendation system: recording that someone is a friend or has liked a post does not itself determine which content appears first. Ranking systems use those connections, together with other inputs, to select and order content for a particular viewer. (usenix.org)
Facebook’s January 2021 account of News Feed ranking described a multistage process rather than a single sorting rule. Candidate selection assembled eligible posts from connected friends, Groups, and Pages, including previously unseen posts and previously viewed posts that had attracted new conversation. Signals included content type, similarity to other items, and the viewer’s interaction history. Machine-learning models predicted different actions, such as liking or commenting; these predictions could disagree about which post should rank highest. Facebook said it combined them into scores informed partly by surveys about meaningful interactions and whether content was worth a person’s time. (about.fb.com)
The same description distinguished preliminary filtering from more computationally expensive ranking. Integrity processes assessed posts, and a lightweight model narrowed the candidate pool to approximately 500 items before more powerful models performed the main scoring. A final contextual pass adjusted the ordering for considerations such as content-type diversity, avoiding an uninterrupted sequence of videos. These figures describe the published 2021 architecture, not fixed parameters for every subsequent Feed implementation. (about.fb.com)
In June 2023, Meta published 22 system cards covering ranking and recommendations across Facebook and Instagram, including Feed, Stories, and Reels. These distinguished content from existing connections from “unconnected” recommendations originating with people, Groups, or accounts a viewer did not follow. Predicted sharing was one example of a relevance estimate; Meta described combining behavioral predictions with survey feedback rather than treating any single action as a complete measure of value. Its documented Facebook controls included “Show more” and “Show less” feedback. Integrity-related signals could also reduce distribution of problematic or low-quality material. These publications describe stages, inputs, and selected predictions, rather than disclose a complete executable algorithm; Meta explicitly withheld some defensive signals to limit circumvention. (about.fb.com)
A dated infrastructure case study is TAO, described by Nathan Bronson and colleagues at the 2013 USENIX Annual Technical Conference. TAO provided typed objects and directed associations, backed by MySQL databases for persistent storage. Its distributed architecture combined database shards with leader and follower cache tiers: followers handled nearby requests and forwarded cache misses and writes to leaders. Regional replication and asynchronous cache-maintenance messages supported geographically dispersed operation, favoring availability and efficiency over strong consistency. The paper reported a deployment spanning thousands of machines, serving many petabytes of data and processing one billion reads and millions of writes per second. These were graph-storage operations, not counts of recommendations or page views. (usenix.org)
Developer interfaces and third-party integration
Facebook’s developer platform enables external applications and websites to connect their services to Facebook accounts and, within permitted limits, obtain information or publish activity. Its application programming interfaces (APIs) define the operations available to developers; Facebook Login provides an account-based entry point without requiring a separate username and password for each service. Integrations have included photo-sharing applications requesting photographs and calendar applications displaying friends’ birthdays. Their scale was substantial: in August 2013, Facebook reported, using internal measurements, more than 850 million Facebook Login uses per month, averaged over the preceding three months. This measured login activity, not unique users or applications. (ftc.gov)
Authorization to sign in, permission to retrieve particular information, and permission to publish back to Facebook were distinct controls. An August 2013 mobile Login update required applications to request posting permission separately, allowing someone to use Facebook for registration without also sharing a playlist or workout with friends. In April 2014, Facebook announced more granular information-sharing controls: a user could, for example, authorize disclosure of an email address while withholding a birthday. It also announced an application dashboard for managing permissions and removing connected apps, alongside a review process for new applications using Login. These controls concerned the categories and operations authorized, rather than a single unrestricted grant of account access. (about.fb.com)
The earlier Graph API model also distinguished imperfectly between an application’s own users and people connected to them. According to the United States Federal Trade Commission’s documented account, Version 1 allowed developers to collect profile information from people who installed an application or interacted with a connected website, as well as information about their Facebook friends. Those friends could have had no direct interaction with the application. Consequently, one person’s authorization could provide a route to information about other people, extending collection beyond the population that had directly chosen to use the service. (ftc.gov)
Facebook announced Graph API Version 2 in April 2014, restricting that route to friends’ profile information. The FTC’s account identifies an important transitional distinction: new applications were automatically subject to the restrictions, while existing applications received one year before the limitations took effect. The announcement therefore did not immediately eliminate Version 1 access. Aleksandr Kogan’s previously registered application was among those grandfathered under the earlier model. Its relevance to developer-platform history is that an application’s registration history determined its available access during the transition; the subsequent Cambridge Analytica data-use controversy involved additional questions of disclosure and privacy. (ftc.gov)
Third-party integration thus depended on both technical access rules and data-governance obligations after information had been obtained. In March 2018, Facebook announced investigations of applications with substantial access under the earlier model, audits where suspicious activity was identified, and platform bans for developers found to have misused personally identifiable information. It also announced plans to terminate information access after three months of application inactivity and require review for additional categories of Login data. Separately, the FTC’s Cambridge Analytica decision established that representations about information collection could be deceptive even where an interface technically enabled the collection: developer access did not remove responsibility for truthful statements about subsequent data practices. (about.fb.com)
Privacy, data practices, and security
Meta’s privacy disclosures distinguish information supplied by account holders, information generated through activity, device-related information, and information received from other businesses. These categories include registration details, published content and communications, content viewed or engaged with, browser and application logs, and device information. Advertisers, application developers, publishers, and other partners can also transmit information through business tools, social plugins, Facebook Login, and application programming interfaces. Collection therefore extends beyond information deliberately published on Facebook: interactions with external services can provide additional signals for advertising and measurement. These disclosures describe collection across the relevant Meta products, rather than a separate Facebook-only dataset. (facebook.com)
Controlling a post’s audience is distinct from controlling the platform’s collection, retention, or use of information. An audience restriction determines who may see material through a particular sharing function; it does not, by itself, determine every subsequent processing operation. This distinction was central to the Federal Trade Commission’s (FTC) 2019 allegations that Facebook had shared information with applications installed by users’ friends despite restrictive privacy settings. The resulting settlement also addressed uses unrelated to publication, prohibiting advertising use of telephone numbers obtained specifically for security purposes and requiring deletion from company servers of personal information deleted by users. These were privacy obligations concerning processing and retention, not simply controls over public visibility. (ftc.gov)
The Cambridge Analytica case involved an application developed by Aleksandr Kogan and repurposed in 2014 to collect Facebook profile information. Facebook’s earlier developer interface enabled applications to obtain information about participating users and their friends, including people who had never interacted with the application. According to the FTC’s account, the application collected Facebook identifiers from at least 250,000 direct participants and identifiers, names, and other information from between 50 million and 65 million friends. Participants had been told that names or other identifiable information would not be downloaded. In its decision announced in December 2019, the FTC found that Cambridge Analytica had violated Section 5 of the FTC Act through deceptive representations and ordered deletion of the deceptively obtained data and associated work product. The findings concerned collection and deception; they did not establish that the data changed an election outcome. (ftc.gov)
On July 24, 2019, the FTC announced a separate settlement under which Facebook would pay a $5 billion civil penalty to resolve allegations that it had violated the commission’s 2012 privacy order. Its legal basis was noncompliance with that earlier order, rather than a determination about Cambridge Analytica’s electoral influence. The settlement established a 20-year oversight framework, including an independent board privacy committee, designated compliance officers, and quarterly privacy-program certifications by chief executive Mark Zuckerberg and the compliance officers. It required independent assessments every two years and documented privacy reviews before introducing new or modified products, services, or practices. Additional data-governance requirements included closer supervision of third-party developers, strict employee-access controls, and a comprehensive information-security program. (ftc.gov)
Security risks include account compromise and unauthorized access, which differ analytically from data processing performed through permitted interfaces but subsequently challenged as deceptive or improper. Meta’s 2023 annual report identified malware, phishing, social engineering, scraping, and hacking as threats, and acknowledged that such activities had occurred on its systems. It also described assessments of third parties with access to company systems or data and contractual requirements for security controls and breach notification. These disclosures cover both attacks against technical infrastructure and attempts to exploit people or legitimate access arrangements; a disputed privacy practice need not involve penetration of a computer system. (sec.gov)
Content governance and platform accountability
Facebook’s Community Standards are platform rules defining permitted content and conduct, distinct from national law and legal orders restricting access to particular material. Meta formulates these rules and translates them into enforcement guidance; in April 2018, Facebook published detailed internal guidelines previously available to content reviewers. Policy development and enforcement are separate activities: rules establish categories and contextual exceptions, while automated systems and human reviewers assess particular posts. Facebook’s November 2019 enforcement report described text and image matching alongside machine-learning classifiers that identified suspected hate speech for review. It also documented automatic removal of some material identical or closely similar to previously established violations, while identifying contextual interpretation as a reason for human involvement. (about.fb.com)
Enforcement measures have different effects. Removal makes an item unavailable; reduced distribution leaves it accessible but lowers its visibility in feeds; warning screens or contextual labels provide information or require an additional step before viewing. Account restrictions affect a person’s ability to use features, rather than merely the availability of one post. Facebook’s February 2023 penalty-system announcement, for example, distinguished educational warnings from restrictions on persistent violators and retained stronger immediate penalties for serious violations. Consequently, counts of removed posts, labeled items and restricted accounts measure different units and interventions; none alone measures how often users encounter prohibited content. Facebook has historically described removal, reduced distribution and contextual information as three distinct approaches rather than interchangeable sanctions. (about.fb.com)
Users can request reconsideration of eligible enforcement decisions. Facebook expanded post-removal appeals in 2018, initially covering nudity or sexual activity, hate speech and graphic violence. After exhausting Meta’s internal appeals process, users may submit eligible content disputes to the Oversight Board, which selects cases for review. Its decisions on individual content cases bind Meta unless implementation would violate the law; broader policy recommendations are non-binding. Published decisions and the board’s recommendation tracker provide scrutiny beyond individual notifications. Meta also publishes quarterly Community Standards Enforcement Reports, including measures of content actioned, proactive detection, appeals, restoration and estimated prevalence. These distinguish enforcement volume from exposure and reversals: a proactive-detection rate concerns action before user reports, whereas prevalence concerns how frequently violating material is seen. (about.fb.com)
On January 7, 2025, Meta’s chief global affairs officer, Joel Kaplan, announced changes intended to reduce enforcement errors. These included concentrating proactive automated detection on illegal and high-severity violations, relying more on user reports for less severe violations, and requiring greater confidence before removal. Meta estimated that 10–20% of its daily content removals in December 2024 might have been mistakes, excluding large-scale adversarial spam actions. This was a company estimate across its platforms, not an independently established Facebook error rate. Likewise, its subsequent claim of roughly 50% fewer enforcement mistakes in the United States between the fourth quarter of 2024 and the first quarter of 2025 was a company-reported assessment. (about.fb.com)
The same January announcement proposed replacing third-party fact-checking with Community Notes, initially in the United States. On March 13, 2025, Meta announced testing beginning March 18 across Facebook, Instagram and Threads; notes would initially remain unpublished. An April 7 update announced the end of US third-party fact-checking and the gradual public appearance of notes. Under the announced design, contributors wrote notes of up to 500 characters with supporting links. Publication required agreement among contributors whose rating histories indicated differing perspectives, rather than a simple majority. Meta stated that attaching a note would not itself reduce distribution, unlike earlier fact-check ratings. The transition was geographically limited: the March announcement retained third-party fact-checking elsewhere until subsequent country-specific launches, rather than announcing its immediate worldwide termination. (about.fb.com)
Social communication, news, and empirical research
Facebook’s friendship connections, comments, messaging, and Groups provide different settings for interpersonal communication: maintaining contact with existing acquaintances, responding to personal updates, and discussing shared interests with people outside an established friendship network. These functions establish opportunities for interaction, not evidence that participation necessarily strengthens relationships or reduces isolation. A concrete example is Black Girls Culinary, which Meta described in October 2022 as using group-linked Community Chats for recipe discussions, including a “Meatless Mondays” topic and audio conversations about cooking. Such examples document how communities used particular tools; they do not independently measure changes in social support or offline relationships. (en.wikipedia.org)
Facebook also distributes news through publisher posts and links, recommendations, and users’ reshares. Research distinguishes content displayed inside Facebook, visits referred to external websites, engagement such as clicks and reactions, and subsequent changes in knowledge or attitudes. Jack Bandy and Nicholas Diakopoulos’s 2023 study analyzed browsing records from 1.3 million United States desktop panelists and 3,247 news and information websites during October–December 2020. Facebook was the fourth-largest external referral source to those publishers and referred 3.1% of all panelists to at least one included website. The study found disproportionate amplification of publishers classified as low-quality, but its referral measurements captured neither all news viewed within Facebook nor mobile consumption; they were not measurements of persuasion. (journals.sagepub.com)
The U.S. 2020 Facebook and Instagram Election Study used randomized experiments to separate particular platform mechanisms from their measured consequences. In an experiment reported in Science in July 2023, Andrew M. Guess and colleagues replaced consenting participants’ algorithmically ranked feeds with reverse-chronological feeds for three months. The intervention reduced time spent and activity on both Facebook and Instagram, while increasing exposure to political content and material from sources classified as untrustworthy. On Facebook, it also reduced exposure to content classified as uncivil or containing slurs and increased exposure to ideologically mixed sources. Despite these changes, the researchers detected no significant effects on political knowledge, issue polarization, or affective polarization—the latter referring to hostility toward opposing political groups—during the study period. (networkscienceinstitute.org)
A separate experiment removed reshared posts from consenting Facebook users’ feeds rather than changing their ordering. Published by Guess and colleagues in Science in 2023, it found that removing reshares reduced exposure to political news, including news from untrustworthy sources, as well as overall clicks, reactions, and clicks on partisan news. Participants also demonstrated lower news knowledge, although the authors qualified how confidently that result could be generalized to all users. The intervention produced no statistically significant changes in political polarization or the individual political attitudes measured. Thus, redistribution affected information exposure and learning without producing detectable changes in those attitudinal outcomes. (csmapnyu.org)
Temporary account deactivation tested a different intervention: withdrawing access to the core service. Hunt Allcott, Matthew Gentzkow, and colleagues reported this experiment in PNAS in May 2024, using analysis samples of 19,857 Facebook users and 15,585 Instagram users. Participants were randomly offered payment to deactivate for six weeks before the election or, in the control condition, one week; the principal comparison therefore concerned five additional weeks without access. Deactivation reduced a political-participation index, primarily through less online participation. Facebook deactivation did not significantly change the overall knowledge index, although secondary analyses suggested reduced general-news knowledge and possibly reduced belief in misinformation. Estimated effects on polarization, perceived election legitimacy, candidate favorability, and voter turnout were close to zero. (doi.org)
These experiments strengthen causal inference for their specified interventions, but participation was voluntary and the deactivation sample required willingness to stop using the service for payment. Individual withdrawal also leaves other users, publishers, and the wider information environment largely intact, unlike a platform-wide change. Restricted access to platform data constrains independent research, while the election-study collaboration provided unusually extensive access under arrangements giving academic authors control over analysis decisions and publication. Findings remain bounded by study duration, self-reported measures, the algorithms operating in late 2020, and the United States election setting; they do not directly estimate cumulative effects over years or effects in other countries. (doi.org)
Law, regulation, and jurisdiction
Legal obligations attach to the corporate entities operating Facebook, rather than to the service’s name alone. On October 28, 2021, Facebook, Inc. changed its corporate name to Meta Platforms, Inc. through an amended certificate of incorporation filed in Delaware; this was a renaming of the corporation, not the creation of a separate legal entity. Meta Platforms Ireland Limited is separately identified in European regulatory decisions concerning the Facebook service. Consequently, identifying the responsible company, the processing or commercial activity involved, and its territorial scope is necessary to distinguish United States enforcement against the parent company from European proceedings involving its Irish subsidiary or both entities. (sec.gov)
Privacy and consumer protection overlap but have different legal foundations. In the United States, the Federal Trade Commission announced a $5 billion settlement with Facebook, Inc. on July 24, 2019, resolving allegations that the company had violated a 2012 FTC order by misleading users about their control over personal information. The legal basis was enforcement of that earlier order, rather than a general adjudication of every disputed Facebook data practice. The settlement combined a civil penalty with corporate privacy oversight and compliance requirements. In Europe, data protection proceedings can instead apply the General Data Protection Regulation directly: Ireland’s Data Protection Commission adopted a decision on May 12, 2023, finding that Meta Platforms Ireland’s Facebook-related transfers of personal data from the European Economic Area to the United States infringed Article 46(1), and imposed a €1.2 billion administrative fine. That decision concerned international transfer safeguards, not the same conduct or legal obligation as the American settlement. (ftc.gov)
Competition law addresses a different set of questions concerning market structure and commercial conduct. On November 14, 2024, the European Commission fined Meta €797.72 million in case AT.40684, Facebook Marketplace, under Article 102 of the Treaty on the Functioning of the European Union and Article 54 of the European Economic Area Agreement. The Commission identified dominant positions in the EEA-wide markets for personal social networking services, including hybrid social media platforms, and online display advertising on social media platforms. It found two abuses: tying Marketplace to Facebook, thereby automatically exposing Facebook users to Marketplace and giving it a distribution advantage; and imposing unfair conditions on competing classified-advertising providers advertising on Facebook or Instagram, allowing Meta to use their advertising-related data for Marketplace’s benefit. The decision required Meta to end the conduct and refrain from equivalent practices. These were findings about specified abuses of market power, not a prohibition of Marketplace itself. (ec.europa.eu)
Content-related obligations constitute another regulatory field. The Digital Services Act establishes requirements concerning notices of illegal content, explanations of moderation decisions, complaint mechanisms, and transparency. For designated very large online platforms, including Facebook, it also requires assessment and mitigation of systemic risks and independent compliance audits. These procedural and institutional duties are distinct from both data-protection rules governing personal information and competition rules governing commercial conduct; they do not make every breach of Facebook’s own content standards a violation of national law. (eur-lex.europa.eu)
The procedural stage also changes the meaning of a reported legal development. The Marketplace fine was a European Commission administrative decision, not a judgment resolving subsequent judicial review. On January 28, 2025, Meta Platforms and Meta Platforms Ireland brought an action before the European Union’s General Court, case T-66/25, seeking annulment of that decision or, alternatively, annulment or reduction of the fine. The published notice records the applicants’ requests and grounds of challenge, not a court endorsement of them. Similarly, the FTC’s July 2019 announcement distinguished the allegations underlying its complaint from the obligations imposed by the settlement. Identifying the authority, jurisdiction, date, legal basis, procedural stage, and remedy therefore prevents allegations, negotiated settlements, administrative findings, and judicial judgments from being treated as interchangeable outcomes. (eur-lex.europa.eu)
References
- Facebooken.wikipedia.org
- Ten Years Later, Facebook’s First Users Look Back at Site’s Earliest Daysthecrimson.com
- How They Got Herethecrimson.com
- Business, Casual.thecrimson.com
- History of Facebooken.wikipedia.org
- Facebook Expansion Enables More People to Connect with Friends in a Trusted Environmentabout.fb.com
- Company Infometa.com
- List of Facebook featuresen.wikipedia.org
- Timeline: Now Available Worldwideabout.fb.com
- Facebook, Inc.: 2012 Form 10-Ksec.gov
- Followfacebook.com
- 22-324 Brief for Petitionerssupremecourt.gov
- Facebook Tipscommunications.tufts.edu
- Invite friends to like or follow a Page or profile on Facebookfacebook.com
- Making Groups Privacy Settings Easier to Understandabout.fb.com
- Difference between a member, participant and visitor in a public Facebook groupfacebook.com
- Add an admin or moderator to your Facebook groupfacebook.com
- How to Grow a Facebook Grouploriballen.com
- Introducing New Features to Facebook Groupsabout.fb.com
- Control who sees or joins your Facebook eventfacebook.com
- Reactions Now Available Globallyabout.fb.com
- More Ways to Share With the Facebook Cameraabout.fb.com
- Connect with Public Figures Through Liveabout.fb.com
- Introducing New Ways to Create, Share and Discover Live Video on Facebookabout.fb.com
- Launching Reels on Facebook in the USabout.fb.com
- Launching Facebook Reels Globally and New Ways for Creators to Make Moneyabout.fb.com
- Making it Easier to Create Videos on Facebookabout.fb.com
- Introducing Home and Feeds on Facebookabout.fb.com
- Search FYI: Find What the World is Saying With Facebook Searchabout.fb.com
- Introducing New Facebook Local Tab, Messenger Communities, AI Integrations and Moreabout.fb.com
- Meta Platforms, Inc. Quarterly Report for June 30, 2023sec.gov
- Meta Platforms, Inc. Annual Report for December 31, 2023sec.gov
- Meta Annual Report 2023s21.q4cdn.com
- Americans’ Social Media Use 2025pewresearch.org
- Social Media Fact Sheetpewresearch.org
- Americans’ Social Media Use 2025: Reportpewresearch.org
- Hard Questions: What Information Do Facebook Advertisers Know About Me?about.fb.com
- Toward fairness in personalized adsabout.fb.com
- Facebook submission in response to ACCC preliminary reportabout.fb.com
- Introducing Marketplace: Buy and Sell With Your Local Communityabout.fb.com
- Commission fines Meta €797.72 million over abusive practices benefitting Facebook Marketplaceec.europa.eu
- TAO: Facebook’s Distributed Data Store for the Social Graphusenix.org
- How Does News Feed Predict What You Want to See?about.fb.com
- How AI Influences What You See on Facebook and Instagramabout.fb.com
- Commission issues opinion in Cambridge Analytica caseftc.gov
- Updates to Facebook Loginabout.fb.com
- Introducing Anonymous Login and an Updated Facebook Loginabout.fb.com
- Commission Opinionftc.gov
- Cracking Down on Platform Abuseabout.fb.com
- Meta United States Regional Privacy Noticefacebook.com
- FTC Imposes $5 Billion Penalty and Sweeping New Privacy Restrictions on Facebookftc.gov
- Fact Sheet on 2019 FTC Order with Facebookftc.gov
- Facebook-NetzDG Transparency Report July 2022about.fb.com
- Publishing Our Internal Enforcement Guidelines and Expanding Our Appeals Processabout.fb.com
- Community Standards Enforcement Report, November 2019 Editionabout.fb.com
- The Three-Part Recipe for Cleaning up Your News Feedabout.fb.com
- How We're Improving Facebook’s Penalty Systemabout.fb.com
- Our Workoversightboard.com
- Oversight Board Bylaws, February 2023oversightboard.com
- Recommendation Trackeroversightboard.com
- Safety Policies - Safety Center | Metameta.com
- More Speech and Fewer Mistakesabout.fb.com
- Testing Begins for Community Notes on Facebook, Instagram and Threadsabout.fb.com
- How do social media feed algorithms affect attitudes and behavior in an election campaign?networkscienceinstitute.org
- Reshares on Social Media Amplify Political News But Do Not Detectably Affect Beliefs or Opinionscsmapnyu.org
- fb-20211028sec.gov
- Inquiry concerning data transfers from the EU/EEA to the US by Meta Platforms Ireland Limited for its Facebook servicedataprotection.ie
- Decision IN-20-8-1 concerning Meta Platforms Ireland Limiteddataprotection.ie
- Commission Decision in Case AT.40684 — Facebook Marketplaceec.europa.eu
- Regulation (EU) 2022/2065 — Digital Services Acteur-lex.europa.eu
- Digital Services Act: Commission designates first set of Very Large Online Platforms and Search Enginesec.europa.eu
- Case T-66/25: Action brought on 28 January 2025 — Meta Platforms and Meta Platforms Ireland v Commissioneur-lex.europa.eu