Meta is preparing a major shift in how it polices content across Facebook, Instagram, and its other apps, leaning more heavily on advanced AI to detect and remove harmful material while scaling back its reliance on third-party moderation vendors.
The company says these upgraded AI systems will take on a bigger share of content enforcement work once they consistently outperform Meta’s current approaches. The focus is on high-risk areas where speed and precision matter most, including content tied to terrorism, child exploitation, illicit drugs, fraud, and scams.
Meta argues the technology is especially well-suited to repetitive and time-intensive moderation tasks, such as reviewing graphic material, and to fast-moving threat categories where bad actors constantly change tactics. According to Meta, the goal is not only to catch more violations, but to do it with fewer mistakes—improving accuracy while reducing what the company describes as “over-enforcement.”
Early testing results are a key part of Meta’s pitch. The company says its AI can spot roughly twice as much violating adult sexual solicitation content compared with traditional review teams, while cutting the error rate by more than 60%. Meta also reports stronger performance against impersonation and account abuse, including fake celebrity accounts and takeover attempts. The AI systems can detect warning signs such as logins from new locations, password changes, and suspicious profile edits—signals that often show up when accounts are being targeted.
Scams are another major target. Meta says these systems can identify and reduce about 5,000 scam attempts per day in which attackers try to trick people into handing over their login credentials.
Despite the greater automation, Meta says people will still play an important role—especially for complex, high-impact decisions. The company states that experts will design, train, oversee, and evaluate its AI systems, and that humans will remain central for the highest-risk judgments, including appeals for disabled accounts and reports that may involve law enforcement.
This AI-driven push arrives during a period of broader change in Meta’s platform policies. Over the past year, Meta has eased some content moderation rules and shifted how it handles disputed information, including ending its third-party fact-checking program in favor of a community-based notes model similar to the one used on X. The company has also adjusted how it handles political content, aiming to give users more personalized control and lifting restrictions around topics it considers part of mainstream discussion.
The timing also matters for another reason: Meta and other major social platforms continue to face legal pressure, including lawsuits that seek to hold social media companies responsible for harms affecting children and younger users. Against that backdrop, stronger automated enforcement—and proof that it works—could become increasingly important to how platforms defend their safety efforts.
Alongside the enforcement update, Meta also announced a new Meta AI support assistant designed to offer users always-on help. The assistant is launching globally, rolling out to Facebook and Instagram on iOS and Android, and appearing inside the Help Center on desktop for both platforms.
With this combined push—more AI moderation, fewer outside vendors, and round-the-clock AI support—Meta is signaling that the next phase of platform safety and user assistance will be built around automation, with human review reserved for the decisions that carry the biggest consequences.
