As the worlds of social media and artificial intelligence collide, Meta’s Oversight Board is rethinking how it can best execute its regulatory role.
Meta’s Oversight Board is an independent, external regulatory body established in 2018 to review content moderation decisions across Facebook, Instagram, and Threads. The 21-member global team acts like a “supreme court” making binding decisions as well as suggestive recommendations when a user appeals Meta’s original decision. Its monthslong, case-based approach alone may not cut it in the age of AI, Sudhir Krishnaswamy, a legal scholar and academic, and the only Indian on the global board, told Rest of World.
“We are always trying to reduce time taken and trying to get more done. Maybe because of the gen-AI space, some of our work would be less individual case-based and more structured. That’s a possibility. And we’re very open to that,” Krishnaswamy said.
Already, seven in 10 social media images are AI-generated using tools like Midjourney or DALL-E, and eight in 10 content recommendations rely on AI. Nearly half of all social media content by businesses will be AI-generated in 2026. Meanwhile, AI tools are leaving behind non-English speakers.
The key question now is whether the Oversight Board has the capacity and regional reach to identify systemic harms at scale and create precedents that actually shifts product and policy decisions, Rachel Adams, founder and CEO of the Global Center on AI Governance and author of The New Empire of AI: The Future of Global Inequality, told Rest of World.
What won’t work is if you see some of the early AI safety boards that some of the big majors set up — they’ve got all American boards. That is not going to work.”
“With AI-generated content and AI-driven enforcement and moderation, the volume, velocity, and cross-language nature of problems the board was established to monitor and conduct oversight over have exploded,” Adams said. “That would require either a larger board, or a stronger surrounding capacity, in terms of research, regional advisory mechanisms, and faster procedures for urgent situations.”
The Oversight Board is not the first line of defense. Moderation across Meta’s platforms happens at both a machine and a human level (and sometimes both). Users who are unhappy with the moderation outcomes can appeal to the independent, external board. Not all of the appeals will be addressed — the board takes on only the cases it believes will have the biggest lasting impact. In addition to the 21 board members, the Oversight Board has staff members from around the world, and it leans on professional translation services and country context briefings to deliver decisions.
The board’s process, given the need for public comment and nuance, draws out over months. Even before AI blew up following ChatGPT’s public launch in late 2022, the board had received millions of case submissions, and was able to handle only a sliver of them.
Krishnaswamy recognizes that “the speed game has to be machined” at Meta. As a result, the board is open to making broader recommendations to fix the gaps in the technology.
Machine moderation is not new — it has been the majority of moderation across all platforms ever since the Oversight Board was founded — but the tools have gotten more sophisticated over time, Krishnaswamy acknowledged. However, when it comes to linguistic, cultural, and political context, these tools underperform outside of the West.
“Machines get better in some policy areas and then get worse at others,” he said, adding that machine moderation models have gotten good at identifying adult nudity, gathering context signals, and implementing at scale, but misinformation, disinformation, and hate speech remain “too complicated” for machines.
Some examples of the board’s global impact:
- The board pressed Meta to end its blanket ban on the word shaheed (martyr), which is sometimes used by extremists to praise or glorify people who have died while committing terrorist acts, because several Arabic, Urdu, Persian, Punjabi, Bengali, and Hindi speakers use it in completely different contexts.
- In a Kenyan case, the board overturned Meta’s decision to remove a comment on politics, citing the use of a slur. The board ruled the term was not derogatory in terms of ethnicity but instead used “to express political criticism.”
- After the board’s intervention, Meta reversed its decision to takedown threats from commenters against authorities in four separate cases — in Ethiopia, Pakistan, Ukraine, and Italy — admitting the comments didn’t violate its violence and incitement policy, and were “rhetorical expressions of criticism, disdain, or disapproval and not credible threats.”
A direct consequence of the board’s criteria to serve underrepresented, marginalized communities is that it takes on cases globally, from sub-Saharan Africa to the Middle East and South Asia. These populations are still underserved by large language models, which are increasingly being used by Meta platforms for speedy content moderation.
“What won’t work is if you see some of the early AI safety boards that some of the big majors set up — they’ve got all American boards. That is not going to work,” Krishnaswamy said. “It’s not going to work in Turkey, it’s not going to work in India, it is not going to work in Somalia.”
The board, which picked up its first case in October 2020, has issued upwards of 200 decisions and 320 recommendations, as per its five-year assessment published in December 2025. Meta has implemented all of its binding decisions and around 75% of its recommendations.
It’s clear when Meta takes down or reinstates posts and accounts based on the board’s decisions, but difficult to track whether Meta adopts other types of recommendations. The board says that its calls for demystification of how certain algorithms work have only been partially addressed.
The complicated cases will arise when the platforms share humans and agents.”
Adams said the board is better positioned than most private governance mechanisms because it has independence and a human-rights frame, but that AI moderation across all social media platforms is “critically uneven” around the world. The global majority often bears the cost of that unevenness, she said, whether through under-enforcement of moderation rules or limited contextual understanding. For instance, according to Aditya Vashistha, head of Cornell University’s Global AI Initiative, AI frequently misses disability-related slurs in Hindi, and Facebook did not implement a hate speech classifier for Bengali — one of the most widely spoken languages in the world — until 2020, “much later than Western languages.”
Increasingly, there’s a meshing of the space between generative AI and social media, which calls for accountability. While agent-only platforms like Moltbot seem radical, social media platforms are already seeing humans, humans assisted by agents, and agents coexisting.
“You’re going to get a kind of motley crew of people on platforms and you will have some degree of difficulty telling what is what. The complicated cases will arise when the platforms share humans and agents,” Krishnaswamy said. “We don’t have a mandate on gen-AI, but it’s something we’re trying to understand.”
Adams said that in addition to lawyers, journalists, and other civil society groups, the board should consider hiring AI evaluation experts; sociolinguists and regional language specialists; child safety, gender-based violence and harassment experts; as well as labor and worker wellbeing expertise — given the continued human cost of content moderation and review pipelines.
