Meta has published adversarial threat reporting since 2017. In 2026 it issued two semiannual editions. The first-half report is labelled March 2026 and says its case studies span work through 2025, with coordinated inauthentic behaviour (CIB) limited to networks removed in the fourth quarter of 2025. The second-half report is labelled August 2026 and says it covers investigations and actions taken primarily in the first half of 2026. Every figure in this briefing is Meta’s own published number or qualitative claim. Nothing here has been re-counted from Facebook or Instagram telemetry.
CIB, in Meta’s policy language, is a network of inauthentic assets that uses false identities as a central method and tries to mislead people or the company. Enforcement is described as behaviour-based, not based on the political content of posts. The March 2026 edition’s CIB section includes an Iran-origin operation aimed at the United States that used Instagram personas Meta described as fake political-scientist “creators” plus amplifier accounts, and a Pakistan-based domestic operation that Meta said used AI-assisted multilingual content and stolen imagery. Beyond those deep dives, Meta wrote that it continued to remove other networks it attributed, as its own assessment, to operators in Russia, China, and Iran. Followers of inauthentic Pages are not the same as members of the operation; the report lists follower counts on assets it removed, which is a reach estimate, not proof of persuasion.
Fraud is the larger volume story. In the March 2026 executive summary and body, Meta wrote that it removed over 10.9 million Facebook and Instagram accounts and 600,000 Facebook Pages associated with scam-centre activity in Southeast Asia and the Middle East during 2025. Between November 2025 and January 2026, it said it removed over 344,000 ads across Facebook and Instagram promoting “nudify” apps and sent cease-and-desist letters to 46 companies. Case studies in that edition include a Cameroon-based animal-rehoming fraud of about 12,000 accounts, a fake-rental network of about 63,000 accounts that Meta said was operated primarily from Bangladesh and Nigeria, and a romance-oriented Bangladesh-based set of about 8,600 accounts and 20 Pages. Those geographies are Meta’s operational labels for where it assessed the operators sat, not a statement about Bangladeshi victims.
The August 2026 report describes further industrialisation. Meta said it took policy-violating action against a coordinated network of more than 500,000 accounts and Pages using AI-generated engagement bait for gift-card fraud; against about 575,000 predominantly dormant accounts mass-produced in Pakistan that it believed were likely warehoused as pre-aged inventory; and against 5,000 Instagram accounts from Uganda impersonating charities. Working with the US Department of Justice, the Royal Thai Police, and industry partners, it said a third coordinated disruption removed more than 1.4 million assets and that shared intelligence contributed to 63 arrests. Those arrests are a law-enforcement outcome Meta associated with the operation; this briefing does not independently verify charging documents.
CIB in the August edition includes new activity Meta attributes to the actors behind Doppelganger, which it calls the most persistent Russian-origin covert influence operation it tracks. It wrote that a high-volume advertising line continued to target France and Germany and expanded to Hungary, with offshoots aimed at Armenia, Moldova, and the United States. It also listed operations it said originated from Russia, France, Spain, Bulgaria, Iran, Israel, and Ukraine, and said its cumulative CIB record includes targeting of at least 135 countries and origins in 81 countries. Attribution in all of these sentences is Meta’s. The same report said a terrorist organisation’s reconstitution attempt led to removal of about 9,000 assets, with additional disruptions of about 2,000 assets in Pakistan and more than 1,000 in Colombia, plus custom-messaging campaigns it said reached about 130,000 younger users in Pakistan and 20,000 in Colombia. Surveillance-for-hire enforcement named Intellexa/Cytrox, Italy-based ASIGINT, two China-based traffic-interception firms, and Israel-based Bindecy as companies Meta said it actioned; those are Meta’s designations.
What this means in practice
Readers in Bangladesh should treat Meta’s Bangladesh-labelled scam case studies as the company’s account of operator location, useful as a warning that fake rental and romance lures are a documented enforcement theme, not as a victim census. Page admins can use the reports’ tradecraft notes — lifestyle bait, off-platform payment, AI personas — to train staff. None of these takedown counts can be used as a promise that a given fraudulent Page is gone or will stay gone. Meta itself wrote in the August edition that disruption displaces threats more often than it eliminates them. Cyber71 cannot file CIB reports as a privileged partner or obtain unlisted indicators from Meta.
What is still unclear
The reports are curated. Meta says they are not the entire universe of inauthentic-behaviour enforcement. Recidivism, false positives, and the share of scam ads that users still see are not independently measurable from the PDFs. Arrest totals, origin countries, and “high confidence” attributions are investigative judgments. Cross-platform presence mentioned in case studies was not fully enumerated in the sections loaded for this item.
Sources
- Meta Transparency Center — Meta H2 2026 Adversarial Threat Report2026-09-24
- Meta Transparency Center — Meta H1 2026 Adversarial Threat Report2026-09-24
Cyber71 links the primary material behind each report so readers can check it directly. External pages are not under our control and may change.
