Meta rolls out its AI support assistant globally and expands automated safety systems

Meta announced on March 19, 2026 that it is broadening how AI is used for both customer support and enforcement across Facebook and Instagram. In markets where Meta AI is already available, the company is rolling out the Meta AI support assistant globally so users can handle common account problems at any time of day without waiting for a human queue.

Support Moves First

Meta says the support assistant can answer questions, reset passwords, manage account settings, and help users report scams, hacked accounts, or impersonation. According to the company, the assistant is available 24/7 and can respond in under five seconds. Meta also said the system speaks languages used by roughly 98% of people online, which suggests the feature is designed as a default support layer rather than a niche add-on for English-speaking markets.

The rollout is also expanding into account-access cases in the United States and Canada, where Meta says AI can already help some users regain access to their accounts. That matters because account recovery is one of the highest-friction support problems on large consumer platforms, and automation can materially cut wait times if it is accurate enough.

Safety Systems Are Also Improving

• Meta said advanced AI systems are now identifying about 5,000 scam attempts per day that were previously not caught.
• The company said reports of celebrity impersonation fell by more than 80% in some areas after those systems were deployed.
• Meta said newer classifiers are finding twice as much violating adult sexual solicitation and twice as many unwanted sexualized comments while reducing certain enforcement mistakes by more than 60%.

The bigger takeaway is that Meta is using AI not only as a visible consumer feature but also as operational infrastructure for trust and safety. That is strategically important because support volume, fraud, and moderation errors all scale with platform size. If Meta can automate a larger share of those workflows without creating new error modes, it lowers service cost while improving response time for billions of users.

The risk is the same one that follows every large automated moderation system: accuracy has to keep improving across languages, edge cases, and appeals. Meta's own framing acknowledges that AI support and AI enforcement are only useful if they speed up help without making the platform feel less accountable. For now, the announcement shows Meta pushing AI deeper into both the front door and the back office of its consumer apps.