Meta Splits Up Its Responsible AI Team
Meta Platforms has split up a team responsible for understanding and preventing harms associated with the artificial intelligence technology it’s developing, as the company diverts more resources to its generative AI work, according to an internal post reviewed by The Information. The changes are part of a wider reshuffle of the AI teams that Meta announced internally this week.
Most employees on the Responsible AI team will move to Meta’s generative AI team, formed in February to make generative AI products. A Meta spokesperson said other members will move to the company’s AI infrastructure unit, which works on the systems and tools to build and operate AI products.
THE TAKEAWAY
• Most employees on Responsible AI will move to generative artificial intelligence unit
• The team was already weakened by layoffs and the loss of a high-profile internal advocate
• Move is part of a broader reshuffle of Meta’s AI teams announced this week
The reorganization follows a series of shake-ups on the Responsible AI team and raises questions about how Meta will ensure the safety of its AI products outside generative AI, including the algorithms that rank and recommend content in Facebook and Instagram feeds and identify harmful posts on the apps. Over the past year, Meta has poured large amounts of money and resources into building AI models to compete with OpenAI and Google. It also has launched generative AI chatbots for Facebook, Instagram and WhatsApp.
“We continue to prioritize and invest in safe and responsible AI development, and these changes will allow us to better scale to meet our future needs,” Meta spokesperson Jon Carvill said in a statement. “While the majority of the RAI team will sit in the Gen AI org, it will continue to support relevant cross-Meta efforts on responsible AI development and use.”
The Responsible AI team has historically worked across Meta’s divisions. Last year, Meta settled claims by the U.S. Department of Justice that the company’s algorithms for determining which Facebook users were shown housing ads improperly relied on characteristics like race and gender. As part of the settlement, Meta agreed to change its ad technology, saying in a blog post, “We continue to pursue work to embed both civil rights considerations and responsible AI into our product development process.”
The Responsible AI team formed in 2019 to ensure that Meta designed and used AI technology fairly and safely. Former Meta executive Joaquin Quiñonero Candela led the effort, which was an outgrowth of the Society and AI Lab, another team Candela built at Meta to help the company develop AI responsibly. Candela left Meta in 2021, costing the Responsible AI team a high-profile advocate who had the ear of CEO Mark Zuckerberg. Other high-level departures in 2022, including that of former Chief Technology Officer Mike Schroepfer—a strong advocate for AI within Meta—and Jérôme Pesenti, former vice president of AI, also created instability.
Last year, Esteban Arcaute took over the Responsible AI team, which moved into Meta’s Social Impact team under former lead Emily Dalton Smith—part of another reshuffle of the AI teams. To some inside Meta, the move indicated that the company considered the Responsible AI team less of a priority, as the reorganization added more layers between it and Zuckerberg, a former employee said. Meta later disbanded another team it had moved into Social Impact—Responsible Innovation—that was tasked with addressing potential harms tied to Meta’s products.
More changes followed. Meta laid off several product designers and user experience researchers in Responsible AI or moved them to other teams amid widespread layoffs over the past year, according to two people with knowledge of the situation. Several senior employees who worked on AI policy also left the company in recent months, including Miranda Bogen, former policy lead, and Kevin Bankston, former policy director. Neither responded to a request for comment.
Even before this week’s reorganization, senior leaders had recently directed members of the Responsible AI team to focus on generative AI and products using the technology, two people with knowledge said. That marked a departure from the team’s charter to research and provide guidance on responsible AI across the company. According to Meta’s website, the core pillars of responsible AI are privacy and safety, fairness and inclusion, robustness and safety, transparency and control, and accountability and governance.