That was Facebook’s motto for its first decade. Meta on the Record collects the court rulings, regulatory findings and reported events that followed, each labeled by strength of evidence and linked to its sources. It also records what Meta did well.
Researched, written and organized with AI (Anthropic’s Claude). Every entry links to its sources. How this was made
This is one way to read the record: five events from Meta’s record on teenagers, in the order they happened. Each links to its entry so you can check it and draw your own conclusion.
Meta’s researchers studied Instagram’s effect on teenagers for at least three years. A 2019 internal presentation said the app made body-image issues worse for one in three teen girls who already had them.
At the same time, Meta was developing Instagram Kids, a version of the app for children under 13. It paused the project in September 2021, after its research became public.
The Wall Street Journal published the research. Frances Haugen, the former employee who took the documents, testified to the Senate weeks later. Meta said the reporting mischaracterized its work.
Reuters obtained Meta’s internal rules for its AI chatbots. They allowed the bots to engage children in romantic or sensual conversations. Meta confirmed the document and removed those passages after Reuters asked about them.
Meta settled with state attorneys general for at least $12.1 billion over ten years, with design changes and an independent auditor. Meta did not admit wrongdoing. That same month, a New Mexico judge declared its platforms a public nuisance.
In 2011, Facebook settled FTC charges that it deceived users about their privacy and agreed to stop. Cambridge Analytica came to light in 2018. In 2019, Facebook paid $5 billion to settle FTC charges that it had violated that order. 2009 entry 2018 entry
Each entry carries one of seven evidence labels and, where available, links to its sources. Open any entry for figures, Meta’s response, and the current status of the case. Entries were researched and drafted with AI; check the linked sources for anything that matters to you.
Entries per year. Select a year to jump to it.
Beacon told users’ friends what they bought or rented on partner websites, and it was switched on unless people opted out. Zuckerberg apologized a month after launch. Facebook shut Beacon down in 2009 to settle a class action.
In their words“We’ve made a lot of mistakes building this feature, but we’ve made even more with how we’ve handled them.” (Mark Zuckerberg, 2007)
According to the FTC, Facebook changed its site so that information users may have marked private, such as friend lists, became public, without warning them or asking first. It also let apps reach nearly all of a user’s personal data. Facebook settled in 2011.
For one week in January 2012, Facebook changed how much positive or negative content people saw in their feeds, then measured whether their own posts turned happier or sadder. They were not told. Facebook’s researchers published the results in 2014.
Bobbi Duncan and Taylor McCormick at UT-Austin had set their privacy controls so their fathers would not learn they were gay. When the Queer Chorus president added them to the choir’s Facebook group, Facebook announced it to all their friends. Any user could be added to a group without approving it.
In their words“Their unfortunate experience reminds us that we must continue our work to empower and educate users about our robust privacy controls.” (Facebook, 2012)
Facebook suspended drag performers and transgender users who did not use their legal names, after one person reported several hundred of their accounts as fake. Chief product officer Chris Cox apologized on October 1, 2014, and Facebook changed how the policy was enforced.
In their wordsCox apologized “for the hardship that we’ve put you through.” (Chris Cox, 2014)
A US lawsuit seeking $150 billion was dismissed. The Ninth Circuit affirmed in April 2026.
As Myanmar’s military drove more than 700,000 Rohingya into Bangladesh, UN investigators said social media, meaning Facebook in Myanmar, played a “determining role” in the hatred. A report Facebook commissioned found it had not done enough. Amnesty International concluded in 2022 that its algorithms substantially contributed.
In their words“Prior to this year, we weren’t doing enough to help prevent our platform from being used to foment division and incite offline violence.” (Facebook, 2018)
Facebook told Congress that Russia’s Internet Research Agency published about 80,000 posts between June 2015 and August 2017, mostly on divisive subjects such as race and guns. About 29 million people were served the posts directly; sharing spread them further.
Rumors and hate speech spread on Facebook ahead of riots in which mosques and Muslim-owned businesses burned and at least three people died. Officials said Facebook had only two people reviewing Sinhala content. The assessment Facebook commissioned found posts may have led to offline violence.
In their words“We recognize, and apologize for, the very real human rights impacts that resulted.” (Facebook, 2020)
The class settlement became final in May 2025. In September 2026, a New Mexico jury found Meta’s promises after the scandal willfully misleading.
A researcher’s quiz app harvested data from up to 87 million users, most of them friends of the people who took the quiz, and passed it to a political consulting firm. Facebook learned of it in 2015, asked for the data to be deleted, and considered the matter closed until news reports in 2018.
Meta’s responseMeta settled the class action without admitting wrongdoing.
Rumors about child kidnappers spread on social media, including WhatsApp, ahead of a wave of mob attacks in India. In every case IndiaSpend reviewed, no evidence of kidnapping was found. In July 2018 WhatsApp began labeling forwarded messages and limited forwarding in India to five chats at once.
Meta said it will appeal.
Attackers chained three bugs in the “View As” feature to steal login tokens. Exposed data included names, contact details, locations, workplaces, religion and posts. In December 2024 Ireland’s regulator fined Meta for failing to build in data protection and for gaps in how it reported the breach.
Since 2016, Facebook had paid people aged 13 to 35 up to $20 a month in gift cards to install a “Research” app with root access to their phone traffic. Some minors only had to tick a box claiming parental consent. Facebook ended the program on iPhone after TechCrunch reported it.
In their words“There was nothing ‘secret’ about this; it was literally called the Facebook Research App.” (Facebook, 2019)
The US housing department charged that Facebook’s ad system let advertisers target and exclude people by protected traits and delivered housing ads in discriminatory ways. In a 2022 settlement with the Justice Department, Meta retired a targeting tool and paid the largest penalty the Fair Housing Act allows.
Contract moderators who spent their shifts reviewing rape, murder and suicide footage sued after developing PTSD. The settlement covered moderators in four states and required counseling and tools such as muted audio by default.
Facebook’s “tag suggestions” feature ran facial recognition on users’ photos by default from 2011. Illinois users sued under the state’s biometric privacy law.
Meta challenged the fine in Ireland’s High Court.
Scrapers abused Facebook’s contact-import tools to collect names, phone numbers, birthdates and locations from users in 106 countries, then posted the set on a hacking forum. Ireland’s regulator found Meta had not built in adequate protection.
Ireland’s regulator found WhatsApp broke EU rules in how it explained sharing people’s data with other Meta companies.
Internal slides said Instagram made body-image issues worse for one in three teen girls who already had them. Among teens who reported suicidal thoughts, 13% of British and 6% of American users traced them to Instagram.
Meta’s responseMeta said the reporting mischaracterized research that, in some cases, relied on input from 40 teens.
Kenya’s High Court ruled in April 2025 that it can hear the case. The claims have not been tried.
Professor Meareg Amare Abrha was killed in Bahir Dar during Ethiopia’s civil war. His son alleges that Facebook posts doxxed him and called for violence, and that Meta did not respond to reports until after his death.
Nebraska police served Meta a warrant for Messenger chats between a mother and her 17-year-old daughter, and the messages became evidence. The mother later pleaded guilty. Meta complied with a valid warrant. The chats were readable because Messenger did not encrypt them by default at the time.
Meta’s responseMeta said the warrants “did not mention abortion at all.”
Class-action lawsuits over the Pixel on health sites are pending.
The Markup found the Meta Pixel on the websites of 33 of Newsweek’s top 100 US hospitals, sending Facebook a packet of data when people booked appointments. Inside the patient portals of seven health systems, it sent details such as medications and conditions. Several hospitals removed it after the report.
Meta sought judicial review in Ireland.
Ireland’s regulator fined Instagram over business-account settings that made teenagers’ phone numbers and email addresses public.
Ireland’s regulator found Meta relied on the wrong legal basis to process personal data for behavioral advertising on Facebook and Instagram.
Human Rights Watch documented security forces in Egypt, Iraq and Jordan creating fake profiles on Grindr and Facebook to trap LGBT people, leading to arrests, abuse and extortion. HRW has pressed Meta on how its platforms respond to this targeting.
Ireland’s regulator ruled that Meta kept moving Europeans’ Facebook data to the US without adequate protection from US surveillance, and ordered the transfers stopped.
Rulings were postponed in February 2026. The case is unresolved.
Moderators hired through the contractor Sama reviewed graphic content for Meta. After mass layoffs, 184 of them sued, alleging trauma, unlawful dismissal and retaliation for organizing. A Kenyan court held that Meta was their principal employer, and an appeals court confirmed in 2024 that Meta can be sued in Kenya.
Meta began blocking news on Facebook and Instagram in Canada on August 1, 2023, in response to the Online News Act. When wildfires forced the evacuation of Yellowknife weeks later, the city told residents it could not share news links on Facebook, and Meta refused requests from officials and broadcasters to lift the block.
Arturo Béjar, who led Facebook’s user-safety work and later consulted for Instagram, told the Senate that Meta’s own user surveys found 13% of Instagram users aged 13 to 15 had received unwanted sexual advances in the previous seven days. He said his 2021 warnings to Zuckerberg went unanswered.
Meta’s responseMeta said “countless people inside and outside of Meta” work to keep young people safe online.
Texas sued under its biometric privacy law over facial recognition that was switched on by default. Texas says it is the largest privacy settlement ever won by a single state.
Meta’s responseMeta did not admit wrongdoing.
After a 38-month investigation, Nigeria’s consumer protection regulator found WhatsApp’s 2021 data policy broke Nigerian consumer and data protection law. A tribunal upheld the fine in April 2025.
In their words“We disagree with both this decision and the fine and will appeal.” (Meta, 2024)
Meta shut down CrowdTangle, the free tool journalists and researchers used to track how content spread, on August 14, 2024. Dozens of groups had asked it to keep the tool running through the US election, and the European Commission demanded answers under the Digital Services Act. Its replacement is limited to academic and nonprofit researchers.
Meta’s responseMeta said its replacement, the Meta Content Library, was built to meet new regulatory requirements while protecting privacy.
Meta said it will appeal.
The European Commission found Meta dominant in social networking and ruled it abused that position by tying Facebook Marketplace to Facebook and by imposing unfair terms on rival classified-ad services that advertise on its platforms.
South Korea’s privacy regulator found Meta analyzed pages users liked and ads they clicked to build ad categories on religion, political views, same-sex relationships and North Korean defector status for about 980,000 Korean Facebook users, and made them available to about 4,000 advertisers without proper consent.
Meta has appealed to India’s Supreme Court.
India’s competition regulator found WhatsApp’s 2021 policy forced users to accept data sharing with Meta. An appeals tribunal upheld the penalty in November 2025.
Meta rewrote its hateful conduct rules and ended third-party fact-checking in the US. The new rules allow posts calling LGBTQ+ people mentally ill or abnormal, and dropped the ban on referring to transgender people as “it.”
In their wordsThe policy now allows “allegations of mental illness or abnormality when based on gender or sexual orientation.” (Meta’s Community Standards)
In Careless People, former public policy director Sarah Wynn-Williams alleges Meta built censorship tools while courting the Chinese market. An emergency arbitrator, citing her severance agreement’s non-disparagement clause, barred her from promoting the book. The order did not address whether her claims were true.
Meta’s responseMeta called the book “false and defamatory.”
The European Commission found that making users either pay for ad-free Facebook and Instagram or consent to tracking did not offer a real option to share less data.
Researchers found the Meta Pixel, embedded on millions of websites, passed browsing identifiers to the Facebook and Instagram apps on Android. For people logged into those apps, that linked their web browsing, including in incognito mode, to their accounts. Meta had used the method since September 2024 and stopped when the research was published.
Meta’s responseMeta said it decided “to pause the feature while we work with Google to resolve the issue.”
Liability found August 1, 2025. Damages have not been set.
Users of the Flo app entered details about their periods, pregnancies and sexual activity after being promised privacy. A federal jury found Meta intentionally recorded that data through code embedded in the app, violating California’s wiretapping law. The other defendants settled before trial.
In their words“The plaintiffs’ claims against Meta are simply false.” (Meta, 2025)
Reuters obtained Meta’s internal rules for its AI chatbots, approved by legal, policy and engineering staff including its chief ethicist. They permitted the bots to engage children in romantic or sensual conversations. Meta confirmed the document and removed those sections after Reuters asked about them.
Meta’s responseMeta said the examples were “erroneous and inconsistent with our policies.”
Former Meta researchers Jason Sattizahn and Cayce Savage told the Senate that children on Meta’s VR platforms faced sexual solicitation and harassment, and that the company had them erase or change research documenting it. Six current and former researchers made similar allegations.
Meta’s responseMeta called the allegations a “predetermined and false narrative.”
Reuters reported that Meta projected about 10% of its 2024 revenue would come from ads for scams and banned goods, and that it banned advertisers only when it was 95% certain of fraud. The reporting won a Pulitzer Prize.
Meta’s responseMeta called the 10% figure an overly inclusive estimate and says it removed 159 million scam ads in 2025.
The FTC argued Meta bought Instagram and WhatsApp to neutralize rivals, citing a 2008 Zuckerberg email saying it is better to buy than compete. The judge ruled the FTC had not shown Meta holds monopoly power today, largely because of competition from TikTok and YouTube.
A Los Angeles jury found Meta and YouTube negligent in designing platforms that a young woman used compulsively from childhood, contributing to her depression and anxiety. The jury assigned Meta 70% of the responsibility.
Meta’s responseMeta said teen mental health is complex and cannot be linked to a single app.
State investigators posing as children on Facebook and Instagram were shown sexual content and solicited by adults. The jury found Meta violated New Mexico’s consumer protection law and imposed the maximum penalty for each violation.
The Ninth Circuit affirmed the dismissal of the Rohingya refugees’ class action, holding that Section 230 shields Meta from liability for the posts.
GLAAD’s Social Media Safety Index scored Instagram 41, Facebook 40 and Threads 39 out of 100, down from 45, 45 and 40 the year before.
Preliminary. A final decision could carry fines of up to 6% of global turnover.
In preliminary findings, the European Commission said Meta did not adequately assess the risks of Instagram’s and Facebook’s addictive design to users’ well-being, including minors’.
In their words“We disagree.” (Meta)
A New Mexico judge ordered a $567 million fund for treatment, screening and prevention, and changes for minors: like counts hidden by default, notifications paused overnight, and no romantic AI chats.
State attorneys general who sued in 2023 settled in the middle of trial. Meta agreed to hard daily time limits for teens, an independent auditor and a ban on deceptive safety claims.
Meta’s responseMeta denied the allegations and admitted no wrongdoing.
Judge Francis Mathew will set the penalty, up to $5,000 per violation.
In a case New Mexico filed in 2021 over Cambridge Analytica, a Santa Fe jury found that 26 of 29 public statements by Meta and its executives about user data, misinformation, hate speech and equal enforcement were willfully deceptive. The jury sided with the state on 31 of 34 questions.
In their words“We disagree with the verdict and will continue to defend ourselves against efforts to distort our record.” (Meta, 2026)
Four of the people behind the entries in this record.
She had set her privacy controls so her father would not learn she was gay. When the president of the university’s Queer Chorus added her to its Facebook group, Facebook announced it to everyone she knew, including him. Anyone could be added to a group without approving it.
She later spoke publicly about the depression that followed. Facebook said her experience reminded it to keep educating users about its privacy controls.
ReportedMore than 700,000 people fled a military campaign of killing, rape and burned villages. For years before it, Facebook was flooded with posts portraying them as invaders. UN investigators said Facebook played a determining role. Facebook’s own commissioned report found it had not done enough.
A US lawsuit seeking $150 billion for survivors was dismissed, and in April 2026 the appeals court agreed, citing the law that shields platforms from liability for users’ posts.
AdmittedA university professor, he was shot outside his home during Ethiopia’s civil war. His son alleges that Facebook posts identified him as Tigrayan, posted where he lived, and called for violence, and that Meta did not respond to reports until after he was killed.
In April 2025, Kenya’s High Court ruled that it can hear the family’s case against Meta. The claims have not yet been tried.
AllegedHired through a contractor, they reviewed videos of killings, abuse and suicide for removal from Meta’s platforms. After mass layoffs in 2023, 184 of them sued, alleging trauma, unlawful dismissal and retaliation for organizing.
A Kenyan court held that Meta was their principal employer, and an appeals court confirmed Meta can be sued in Kenya. The case is still unresolved.
AllegedMeta’s fines and settlements since 2009 add up to more than $25 billion. Meta’s revenue in 2025 was $201 billion.
The total is a minimum. It leaves out New Mexico’s $942 million judgment and other verdicts under appeal, damages still to be set in the Flo case, and more than 200,000 pending individual claims. Revenue figure as reported by PBS NewsHour.
Things Meta did well, each listed with relevant context.
On the day Facebook let people share their donor status and linked them to state registries, new online registrations jumped 21-fold over the daily average, according to a Johns Hopkins study.
ContextThe surge faded within about two weeks, and registrations do not measure transplants.
Safety Check lets people tell family and friends they are safe during a disaster. After the 2015 Paris attacks, 4.1 million people marked themselves safe within a day, and 360 million people were notified.
ContextFacebook chose when to switch it on, and was criticized for using it after Paris but not after the Beirut bombing a day earlier.
WhatsApp turned on end-to-end encryption for every message and call by default, using the open Signal Protocol.
ContextMeta still collects metadata such as who talks to whom, and encryption makes child-abuse material harder to detect.
Meta commissioned and published outside assessments of its role in Myanmar, Sri Lanka, Indonesia, Cambodia and Israel-Palestine.
ContextEach came after the harm, and Meta paid for them.
Facebook shut down its face-recognition system and said it would delete more than a billion face templates.
ContextIt came amid biometric lawsuits that have since cost Meta more than $2 billion.
Meta built PyTorch, one of the most widely used tools for building AI, and gave it to the Linux Foundation. More than 18,000 organizations had adopted it.
ContextMeta’s Llama models are often called open source. Their license restrictions fall short of the Open Source Definition.
Meta says more than $7 billion had been raised for nonprofits through Facebook and Instagram, with birthday fundraisers a large share.
ContextThis is Meta’s own figure, and it has since scaled back some fundraising features.
Messenger began encrypting personal chats and calls by default for its more than one billion users. An EFF researcher called the move wholly a net positive.
ContextThe default arrived in December 2023, after the Nebraska case in 2022.
A record of the court rulings, regulatory findings and reported events involving Meta, each labeled by how strong the evidence is and linked to its sources.
Cases Meta won are included, and so is what it got right. The labels describe how firmly each entry is established. What you make of the record is up to you.
The entries were researched and written with AI. How this was made explains what that involved.
This site was researched, written, organized and built with AI, using Anthropic’s Claude, at the direction of its editor.
Claude searched news reports, court and regulator announcements, and Meta’s own statements. It drafted each entry’s summary, assigned the evidence labels, checked the figures against the linked sources, and wrote the site’s design and code.
A verification pass on September 28, 2026 rechecked flagged entries against multiple outlets and corrected four errors in quotes and wording. For recent court cases, the sources are news reports of the rulings; links to the court documents themselves are being added over time.
AI tools make mistakes. Every entry links to its sources so you can check it. Send corrections with a source.
Each entry has one label for its strongest evidence. A filled mark means a court or regulator ruled against Meta; outlined and dashed marks mean weaker evidence.
The internal slide was about girls who already had body-image issues. Among them, 32% said Instagram made those feelings worse.
That is Facebook’s upper estimate, and most were friends of the roughly 300,000 people who took the quiz.
Facebook said that many may have been served the posts. It could not say how many actually saw them.
It was an internal projection for 2024 that also counted ads for banned goods. Meta disputes it.
Myanmar’s military carried out the atrocities. UN investigators and Amnesty found Facebook played a determining or substantial role in spreading hate. US courts dismissed the lawsuit on legal grounds without ruling on those facts.
Meta complied with a valid warrant that did not mention abortion. The messages were readable because they were not end-to-end encrypted.
The FTC lost. The judge ruled it had not shown Meta holds monopoly power today.
Reports of the 2026 state settlement range from $12.1 billion to about $18 billion depending on what is counted. This site uses the guaranteed minimum.
Day-one organ donor sign-ups appear as 13,012 and 13,054 in the same study. Nepal Safety Check figures appear as 7 and 8.5 million. The Flo verdict is dated August 1, 2025; later dates are reporting dates. The September 2026 New Mexico verdict is counted as 43,899,720 violations on the verdict form and 43,899,725 by the state.
Three ways to limit what Meta collects about you.
Accounts Center lets you export your information from Facebook and Instagram. Reviewing it can help you decide what to delete.
Websites and apps report your visits and purchases back to Meta through tracking code. Your account settings let you review and disconnect that history.
Use end-to-end encrypted chats for anything about health, legal matters, or your identity. In the Nebraska case, unencrypted messages were used as evidence.