Meta is driving an infrastructure-centric strategy to lead enterprise AI through strategic acquisitions and internal foundational models, aiming to become the backbone for a new generation of workplace AI agents.

A serious internal AI agent malfunction triggered a Sev 1 security incident, exposing confidential data to unauthorized engineers and prompting a critical governance and safety review for enterprise-scale autonomous systems.

Key questions now center on how to integrate Manus—whether as an independent unit or fully merged—and how governance and security oversight will evolve to support mass adoption without repeating past breaches.

Meta foresees monetization from embedding autonomous assistants in core products, with Manus delivering immediate scale and a pathway to premium business offerings.

The strategy blends acquisitions with building internal capabilities, using Manus to jumpstart an agent platform while navigating governance challenges that could impact trust, privacy, and regulatory scrutiny.

Mark Zuckerberg is personally testing Meta’s internal AI agents, highlighting the shift from a single-user assistant to enterprise-grade agents and the gap between home and workplace AI.

Meta plans to release foundational models Mango (image and video) and Avocado (text) in early 2026, led by Alexandr Wang, to boost visual interpretation and autonomous operation within its agent stack.

Meta’s acquisitions of Moltbook and Manus—reaches of $125 million annual run rate for Manus—are accelerating internal development and providing scalable agent capabilities.

Overall, Meta aims to build the infrastructure for agentic AI faster than rivals, but it must tackle governance and security weaknesses to sustain leadership.