Integrity Reports, H2 2026

UPDATED AUG 27, 2026
We're publishing our semiannual integrity reports, including the H1 2026 Community Standards Enforcement Report, our latest Adversarial Threat Report, the H1 2026 Widely Viewed Content Report, the Transparency Report consisting of H2 2025 Government Requests for User Data, and our Oversight Board Report for H1 2026. The H1 2026 Content Restrictions Based on Local Law will be published later this quarter. All of the reports are available in our Transparency Center.
Some highlights of this report include:
Community Standards Enforcement Report
  • Last year we began providing new data on our global enforcement precision as a measure of over-enforcement. At the end of H1 2026, global enforcement precision averaged around 91% on Facebook and around 92% on Instagram, which was generally consistent with our last reporting.
  • Prevalence of violating content remained consistent across most problem areas, with two exceptions:
    • On Facebook, prevalence for adult nudity and sexual activity decreased (0.10 - 0.11% in Q4 2025 to 0.07 - 0.08% in Q2 2026) by deploying LLMs to close enforcement gaps.
    • Prevalence of violent and graphic content increased by 0.05% on both Facebook (0.15% - 0.16% in Q4 2025 to 0.20% - 0.21% during Q2) and on Instagram (0.09% to 0.10% in Q4 2025 to 0.14% - 0.15% during Q2) due to real world events.
  • We also continue to share data on our efforts to combat fraud and scams on our platforms. So far this year we’ve removed 65 million scam ads from Facebook and Instagram, with 94% of those ads removed before anyone reported them to us.
  • In 2025 we announced that our automated and AI-assisted proactive enforcement systems would focus more on illegal and high-severity violations. Though we have been enforcing violations by designated criminal organizations for some time, this is the first report to include that specific data. Our proactive enforcement rate on violations by these organizations was 98.1% on Facebook and 99.6% on Instagram.
  • Furthermore, we continue to test more advanced Al systems to deliver better enforcement. Since March, initial tests have shown that on average LLMs make 13% fewer mistakes than humans when enforcing against violating content while finding 10% more actual violations.
You can read the full report here.
Adversarial Threat Report
This report includes new insights on attempted misuse of AI to conduct influence operations, offensive cyber activity, and fraud and scams.
AI technologies are now present in the operations of virtually every Coordinated Inauthentic Behavior (CIB) network we disrupt. We're sharing new research into the removal of a network originating in Iran that targeted audiences in the United States with political content. This network included 4 Facebook accounts and 31 Instagram accounts, and amassed more than 79,000 followers. Despite sophisticated operational security, including the exclusive use of U.S. and Canadian proxy infrastructure and personas posing as American activists and students, we were able to discover the network through our own proactive internal detection and attribute it to actors in Iran. Some content produced by this network was AI-generated.
We also detail new activity by the actors behind Doppelganger, the most persistent Russian-origin covert influence operation we track and combat using continually enhanced automated detection and campaign-specific defenses. These actors ran a series of tactically distinct campaigns, including new efforts targeting European elections, and some content produced was AI-generated. The report also provides case summaries for additional CIB operations originating in Russia, as well as operations originating in France, Israel, and Ukraine.
Additionally, we are sharing updates on our enforcement against terrorist organizations, including Strategic Network Disruptions targeting an extreme far-left terrorist organization in Colombia, where we disabled more than 1,000 assets and deployed AI-driven detection systems to maintain continuous enforcement pressure on the network's propaganda infrastructure.
Lastly, we continue to take action to combat fraud and scams on our platforms through cross-sector collaboration and our Fraud Attack Chain framework. In June, Meta helped convene the largest joint disruption operation targeting criminal scam networks to date conducted with the U.S. Department of Justice's Scam Center Strike Force. Working alongside law enforcement agencies including the FBI and the Royal Thai Police and together with cross-industry partners, we removed more than 1.4 million accounts, Pages, and Groups across Facebook and Instagram and shared intelligence that contributed to 63 arrests.
You can read the full report here.
Widely Viewed Content Report
In H1 2026, the top 20 widely viewed domains collectively accounted for about 0.3% of all Feed content views overall, remaining consistent with Q4 2025.
You can read the full report here.
Government Requests for User Data
During the second half of 2025, global government requests for user data increased by 11.1% from 374,516 to 415,977.
In the US, we received 80,009 requests in the second half of 2025, an increase of 1.3%, 78.1% of which include non-disclosure orders prohibiting Meta from notifying the target user. Emergency requests accounted for 6.2% of the total request in the U.S. Meta supports the NDO Fairness Act (H.R. 6048, S. 3663), which would improve the judicial review of NDO requests. We are grateful to the sponsors and supporters of this important legislation, and urge Congress to Act to pass this bill.
Additionally, as a result of transparency updates introduced in the 2015 USA Freedom Act, the U.S. government lifted non-disclosure orders on 11 National Security Letters (“NSLs”). These NSLs, along with the U.S. government’s authorization letters, are available here.
You can read the full report here.
Oversight Board Report
We publicly respond to every Oversight Board recommendation. To date, we’ve committed to implementing or exploring the feasibility of implementing 80% of their 347 recommendations.
During H1 2026, our work with the Board drove progress in several areas:
  • Account-level enforcement: For the first time, the Board reviewed an account takedown decision, extending its impact beyond individual pieces of content to whether removing an account’s presence on our platforms was proportionate.
  • Community Notes: We asked the Board for a Policy Advisory Opinion on how to decide which countries to include in the global Community Notes rollout, seeking their external input before deployment rather than after.
You can read the full report here.
NCMEC CyberTips
As part of our ongoing work to provide young people with safe, positive online experiences, we continue to provide transparency into our efforts to find and report child exploitation to the National Center for Missing and Exploited Children (NCMEC).
In H1 2026, Facebook, Instagram, and Threads sent over 9.9 million CyberTip reports for child sexual exploitation.
Of these reports:
  • Over 1.3 million involved inappropriate interactions with children. CyberTips relating to inappropriate interactions with children may include an adult soliciting child sexual abuse material (CSAM) directly from a minor, online enticement of a minor, minor sex trafficking, or attempting to meet and cause harm to a child in person. These CyberTips also include cases where a child is in apparent imminent danger.
  • Over 8.5 million reports related to shared or re-shared photos and videos that contain CSAM.
Other Integrity Updates
  • We're publishing a new Transparency Center page that provides an overview of how we work with other companies, industry coalitions, and stakeholders to combat harm across the internet through signal sharing, technology, and collective action. Threat actors don't limit their activities to a single platform — they shift across services, sectors, and geographies to avoid detection — and no single company can stop them alone. The page details active signal-sharing programs with participating organizations covering child safety, terrorism and violent extremism, fraud and scams, intimate imagery, self-harm and suicide, and drug harms. It also highlights the open-source technology we've built and made freely available to platforms of all sizes, including ThreatExchange, PDQ photo and video hashing, and Hasher-Matcher-Actioner. You can find the full page here.
  • Today we also published Meta's fifth annual Human Rights Report, covering insights and actions from our work managing those risks globally at Meta in 2025.