Meta split up its Responsible AI team, moving most of its members to the generative AI team formed in February 2023, and the others to an AI infrastructure unit
Meta Platforms has split up a team responsible for understanding and preventing harms associated with the artificial intelligence …
Context & Ripple Effects
Meta had already placed its Responsible AI group under Social Impact in a broad 2022 AI reorganization, then shifted its broader AI effort toward usable generative products in 2023. This move takes the next step: responsibility for AI harms is no longer organized as a standalone team.
Later reorganizations continued to separate product, foundation-model, and research work, including a 2025 split between AI products and AGI Foundations. The 2023 change is an early indicator that Meta was aligning safety expertise with model development and infrastructure rather than keeping it centralized.
First-order effects
- Most Responsible AI staff move into Meta’s generative-AI organization, giving the product-and-model effort direct access to people focused on identifying and mitigating AI harms.
- The remaining staff move to AI infrastructure, ending the team’s standalone structure and distributing its remit across two operating units.
Second-order effects
- Safety review and mitigation work can be incorporated earlier into generative-AI development and infrastructure decisions, but coordination now depends on the priorities and leadership of those separate groups.
- The organizational change reinforces Meta’s 2023 push to redirect AI resources toward usable products, described as a refocus from long-running research toward generative AI.
Third-order effects
- If this operating model persists, AI governance at Meta becomes an embedded engineering and product function rather than an independent organizational checkpoint—potentially improving implementation speed while making cross-company accountability harder to isolate.
- The later recurrence of AI restructurings suggests that organizational design itself is becoming a competitive tool as large labs balance product delivery, model development, research, and compute.
The trend: Big AI developers are folding responsible-AI work into the teams building models and infrastructure as governance shifts from a standalone mandate to an operational capability.