Meta Expands AI Scam Defenses Across WhatsApp, Facebook, and Messenger

Meta is widening its anti-scam response from warnings to AI detection to enforcement

On March 11, 2026, Meta announced a new set of anti-scam tools across WhatsApp, Facebook, and Messenger, along with broader use of AI to detect scam activity. The rollout includes a device-linking warning in WhatsApp, suspicious friend-request alerts in Facebook, and expanded AI scam review in Messenger chats. Meta also said it is using more advanced AI systems to identify celeb-bait, brand impersonation, and deceptive links.

The larger significance is that Meta is no longer framing scam prevention as a single moderation problem. The company is building three layers at once: product warnings that interrupt suspicious actions before they succeed, AI systems that analyze patterns across text, images, and context, and large-scale enforcement against scam ads and scam-center-linked accounts. That is a more operational approach to fraud: reduce successful attacks not only by deleting bad content after the fact, but by adding friction earlier in the user journey.

Each app is getting a different defensive layer

In WhatsApp, Meta said the new warning system uses behavioral signals to flag suspicious device-linking requests. That targets a common takeover path in which a scammer tricks someone into sharing a linking code or scanning a QR code that connects the victim’s account to the scammer’s device. In Facebook, Meta is testing alerts around suspicious friend requests, including cases where an account has few mutual friends or appears to list a different country location. In Messenger, the company is expanding advanced scam detection to more countries, warning users when new-contact chats show common scam patterns such as suspicious job offers and offering an AI-based review of recent messages.

That product-specific design matters. Scam behavior does not manifest in the same way across every Meta surface. WhatsApp faces account-linking abuse, Facebook deals more directly with social-graph manipulation, and Messenger is a natural venue for conversational fraud. Meta’s rollout suggests the company is designing defensive flows around those different attack patterns instead of forcing a single generic safety prompt everywhere.

AI is moving deeper into impersonation and scam detection

Meta said its advanced AI systems analyze multiple signals, including text, images, and surrounding context, to detect more sophisticated scam patterns at scale. The company specifically called out impersonation involving celebrities, public figures, and brands, as well as deceptive links and domain impersonation. Those are areas where rule-based systems often struggle because the fraud may rely on subtle framing, fake social proof, or visual similarity rather than a single obvious indicator.

Meta is also tightening the business side of scam control. The company said it is expanding advertiser verification with a goal of having verified advertisers drive 90% of ad revenue by the end of 2026, up from 70% today. That is an attempt to make fraudulent advertiser identity harder to scale, not just to catch bad ads after they are already running.

The enforcement numbers show the scale of the problem

Meta said it removed more than 159 million scam ads globally in 2025, with 92% taken down proactively before anyone reported them. It also took down 10.9 million accounts on Facebook and Instagram associated with criminal scam centers. In a recent operation with global law enforcement, Meta investigators disabled more than 150,000 accounts associated with scam-center networks in Southeast Asia.

That scale matters because online scams now operate across messaging, social media, dating apps, and crypto channels as a cross-platform criminal industry. Meta’s update shows how large platforms are being pushed toward a combined model of product warnings, AI moderation, advertiser controls, and law-enforcement coordination. The next questions are practical ones: whether these warning flows measurably reduce victimization, how accurate AI scam review proves in real use, and whether advertiser verification changes the economics for fraud networks operating on large ad systems.

Source: Meta