Meta has announced a plan to change how it handles content moderation on its social platforms by relying more on artificial intelligence. The company also introduced a Meta AI assistant to help people regain access to their Facebook and Instagram accounts.

What is changing

Over the coming years, Meta will reduce the number of external human moderators it uses and move monitoring tasks to systems based on large language models. The company says it will depend less on third party vendors and invest more in its internal systems.

Claims about language and nuance

Meta states its AI can understand roughly 98 percent of the languages used online, compared with the company's previous coverage of about 80 languages. According to Meta, these systems can also handle slang and language used in subcultures, which are often hard for earlier tools to interpret.

How this affects safety features

The company plans to use AI to try to make the platforms safer in several ways:

  • Reduce scams that trick people into revealing login details.
  • Stop fake celebrity accounts from spreading.
  • Speed up detection of posts that violate rules on sexual content involving minors.

The role of human reviewers

Meta says it will not eliminate human reviewers, but it will change what they do. Remaining human staff will focus on designing, training, and evaluating the AI systems. They will also handle the most sensitive and high-risk decisions, such as account deactivation appeals and reports to law enforcement.

Meta AI account support

Alongside the moderation changes, Meta is rolling out an AI-powered support feature to help people access their Facebook and Instagram accounts. The tool will operate 24 hours a day and is launching first in the United States and Canada. Meta says it will expand the service to more countries and to cover more account-access problems over time.

The company frames the move as a shift toward internal AI tools and a different mix of human and machine work, with humans concentrating on the highest-risk decisions and on improving the AI itself.