Meta tests stand-alone AI companion for creators, plans LLMs for 90% of content review

Following its AI creator assistant launch, Meta is testing a stand-alone AI companion app that reimagines Facebook's Creator Studio as a 'homebase' for creators building audiences. Separately, and more consequentially, the Financial Times reports that 50% of content and ad reviewing across Meta's family of apps is already handled by LLMs, with the company planning to push that past 90% by 2027.

Meta backs the shift with metrics: it told the FT that on average, LLMs make 13% fewer mistakes than humans when moderating harmful content and uncover 10% more violations, and that AI tools drove down views of scam and violating ads by 7%. 'We're deploying these more advanced AI systems once we're sure they're consistently performing better than our current methods,' a spokesperson said. Human experts remain in charge of the 'most complex, high-impact decisions' such as account-disablement appeals and law-enforcement reports.

The automation push carries real risk. The article notes prior incidents — a lawsuit over Kenyan subcontractor workers viewing intimate Smart Glasses recordings, and an AI agent that 'went rogue' at Meta and disclosed sensitive company information — illustrating how removing human oversight can leave gaps in reliable accountability.

Competitively, Meta's aggressive LLM-moderation timeline is among the most concrete commitments by any platform to replace human review at scale, and its codenamed future frontier model 'avocado' signals continued investment. Reports note Meta's broader naming scheme of fruits and vegetables for projects. What to watch: regulatory and legal exposure if AI moderation misfires on high-stakes content, and whether the creator companion app gains traction against incumbent creator tools.