Meta to Deploy New Content Moderation System by March 1, Documents Show DHS Coordination

The March 1 Rollout

Meta will deploy a new automated content moderation system across Facebook, Instagram, and Threads by March 1, 2026, according to two employees at the company familiar with the project. The system, internally named "Sentinel Shield," will replace the company's current "Cross-Check" program and will use machine learning models to flag posts for review before human moderators examine them, the employees said.

A 47-page internal planning document dated January 15, 2026, outlines the rollout schedule and lists the Department of Homeland Security as a "key external stakeholder" in the project, according to one of the employees, who reviewed the document. The document identifies weekly coordination calls on Tuesdays at 2:00 p.m. Eastern Time between Meta policy staff and officials from DHS's Office of Intelligence and Analysis, the employee said.

The new system will prioritize enforcement against content related to election administration, immigration enforcement, and what the document calls "coordinated influence operations," the employees said. The document specifies that Sentinel Shield will begin testing with 5 percent of U.S. users on February 9, 2026, and will reach all domestic users by March 1, the employees said.

The employees said the system will introduce a new "velocity score" that measures how quickly a post spreads across the platforms. Posts that reach more than 10,000 impressions within 60 minutes will automatically trigger a secondary review queue, the employees said. The threshold was selected after internal testing in December 2025 showed that high-velocity posts accounted for a disproportionate share of content that later received enforcement action, one employee said.

Internal Documents and Government Coordination

A former content moderator who worked on Meta's trust and safety team until December 2025 said the company had begun sharing "trending topic reports" with federal officials in late 2024 but that the January documents show a deeper operational relationship than previously disclosed. The former moderator reviewed drafts of the coordination protocols while still employed at the company.

The planning document references a $3.2 million contract for "government liaison services" that Meta signed with an outside consulting firm in November 2025, according to the two employees familiar with the project. The contract runs through September 30, 2026, and requires the firm to prepare weekly briefings for DHS officials based on Meta's internal data, the employees said. The consulting firm is not named in the document reviewed by the employees, who declined to identify it.

A congressional investigator who has reviewed similar documents obtained through oversight requests said the Tuesday coordination calls appear to go beyond the public meetings Meta has disclosed in its transparency reports. The investigator, who works for a House committee, said staff members are preparing a letter to Meta requesting copies of the agendas and attendee lists for the calls. The letter is expected to be sent by close of business on January 26, the investigator said.

"The document describes an operational pipeline, not just policy consultation," the former content moderator said. "Moderation decisions are being shaped by people who do not work for the company."

The former moderator said Sentinel Shield uses a scoring model trained on more than 14 million previously moderated posts. The model will assign each flagged post a risk rating from 1 to 100, with posts scoring above 75 routed to a special queue reviewed by senior policy staff, the former moderator said. Posts scoring above 90 are flagged for inclusion in the weekly DHS briefing, according to the former moderator.

The employees said the first Sentinel Shield training session for policy staff was held on January 13, 2026, at Meta's Menlo Park headquarters. Approximately 120 employees attended the session, which included a demonstration of the scoring model and a walkthrough of the DHS briefing template, the employees said.

Legal and Political Stakes

Meta has not publicly announced Sentinel Shield. The company's most recent transparency report, published in December 2025, described only routine updates to its existing enforcement tools and made no mention of the new system or the DHS coordination calls.

A lawyer involved in a pending First Amendment case against Meta in the Northern District of Texas said attorneys for the plaintiffs plan to file a supplemental declaration by January 29 citing the internal documents. The case, filed in August 2025, alleges that Meta improperly restricted content from conservative news outlets and commentators during the 2024 election cycle. The lawyer said the documents could support an argument that the company acted as a state actor when making moderation decisions.

Three congressional aides briefed on the plan said Republican staff on the House Judiciary Committee are preparing a hearing for early February focused on social media companies and government coordination. The aides said committee chairman Jim Jordan's office has requested copies of the Sentinel Shield planning document and the consulting contract by January 30.

Watch for three developments in the next 48 to 72 hours: a possible statement from Meta's communications team denying any new government coordination, a response from DHS about the scope of its engagement with social media companies, and the House committee letter requesting documents. The formal Sentinel Shield launch is scheduled for March 1, but the February 9 pilot will be the first visible sign of the system for users.