Oversight Board Newsletter - September 2024: AI Paper, Venezuela decision and more

Hello and welcome to Across the Board, the Oversight Board's monthly newsletter filled with updates about our work and decisions.

In the last month: We released a report on AI and content moderation and published new case announcements and decisions, including an expedited decision about post-election violence in Venezuela. In this edition, we also announce new funding for the Oversight Board.?

Thanks for reading,

The Oversight Board

?? New paper: Content Moderation in a New Era for AI and Automation

This week we released a new paper sharing key recommendations on AI that urge platforms to prepare for a new era of content moderation. Social media has evolved dramatically since Facebook's launch in 2004, with automated systems now deciding what content to show or remove, often without human oversight. These algorithms shape user experiences but can also reinforce biases and interfere with freedom of expression. The Board reviews Meta’s AI-driven content moderation to address these issues. Our work highlights the importance of improving automated systems to balance human rights with scalable tech solutions. Our paper is a call for more responsible, transparent AI in the future of social media content moderation.? Key lessons:

• Platforms should focus their policies on identifying lack of consent among those targeted by the proliferation of non-consensual deepfake intimate images.
• Platforms should use automation to help users understand why their content was removed and provide clear notifications. Users deserve to know if a human or AI made the decision and should be able to appeal with added context.
• The benefits of new generative AI models should be shared fairly across all social media users worldwide.
• Automated moderation systems need ongoing, thorough evaluation to ensure they work effectively for the most vulnerable and at-risk users.
• AI-powered content moderation tools should be developed with input from global experts in human rights, freedom of expression and ethics. Their recommendations for safety should be built into the design.
• Transparency is paramount – third-party researchers worldwide should have access to data in order to assess these tools.
• Platforms should also label content that’s been significantly altered and could mislead users.

https://www.oversightboard.com/news/content-moderation-in-a-new-era-for-ai-and-automation/ ??

???Oversight Board funding assured for next three-year period??

The Oversight Board Trust has confirmed that Meta will provide its next round of top-up funding to the Board.??

This investment, along with current funding, means the Board has guaranteed funding of at least $35 million for each of the next three years. This top-up funding will go into the Board’s Irrevocable Trust, thus protecting the Oversight Board’s independence from Meta.??

The impact of the Board’s decisions and recommendations are clear: Meta’s content policies are being applied more consistently and transparently, ultimately impacting billions of Meta’s users.???

??? The Board issues expedited decision on post-election violence in Venezuela

?Last week, the Oversight Board issued an expedited decision about two cases that involved videos posted after the July 2024 presidential election and during the ongoing protests that followed, referencing state-supported armed groups known as “colectivos,” involved in the crackdown on protesters.

?? Case 1: The Board agrees with Meta’s decision to keep a post on Instagram, showing a group of armed men on motorbikes pulling up to an apartment complex. A woman can be heard shouting that the colectivos are trying to enter the building. The person filming shouts “Go to hell! I hope they kill you all!” Meta found this content did not violate its Violence and Incitement policy because the expression was a conditional or aspirational statement against a violent actor rather than a call to action.

?? Case 2: The Board disagrees with Meta’s decision to remove a video showing a group of men on motorbikes, presumably colectivos, and people running on the street. The man filming shouts that the colectivos are attacking them. The video has a caption in Spanish calling out the security forces for not defending the people and saying that the security forces should go and “kill those damn colectivos.”

Given the context—where security forces are linked to the colectivos and involved in repressing the opposition—the user’s post is better understood as a desperate cry of frustration and fear, not a credible threat.

The Board finds neither post violates Meta’s content policies and, in the context of the ongoing crisis in Venezuela, allowing both pieces of content is consistent with Meta’s values and human rights responsibilities.

Board findings and recommendations

The Board is deeply concerned that Meta’s policy of limiting the spread of political content could prevent users who are expressing political dissent or raising awareness about the situation in Venezuela from reaching a broad audience.

The Board suggests that Meta's Crisis Policy Protocol should include a rule to ensure political content, especially around elections and protests, can reach as many people as non-political content.

The Board also emphasizes that it’s crucial to consider the context to protect political speech, especially in countries facing conflict or restrictions on free speech. When Meta declares a crisis, it should use the Crisis Policy Protocol to better understand how people experiencing state-backed violence communicate on its platforms, like in Venezuela.

In situations where democratic dissent is being suppressed, the threats appear non-credible and the risk of online speech leading to real-world violence is low, Meta should adjust its policies and enforcement accordingly. This approach should be regularly reviewed with input from relevant stakeholders.

https://www.oversightboard.com/decision/bun-xgwxj6hs/

???Update on our cases and decisions

New case announcements:

• We?announced new cases related to the debate on gender identity. The Board selected these cases to assess whether Meta’s approach to moderating discussions around gender identity respects users’ freedom of expression and the rights of transgender and non-binary people.

New case decisions:

• We upheld Meta's decisions to allow three Facebook posts containing the phrase “From the River to the Sea.” Our decision set out that the content did not break Meta’s rules on Hate Speech, Violence and Incitement or Dangerous Organizations and Individuals.
• We overturned Meta's decision to remove a reply to a Threads post criticizing outgoing Japanese Prime Minister, Fumio Kishida, by calling attention to alleged corruption. The Board found the phrase "drop dead/die" (from the Japanese "死ね") wasn't a credible threat and didn’t break Meta’s Violence and Incitement rules.?
• We upheld Meta’s decision to remove a post accusing a political candidate in Pakistan of blasphemy, ahead of the 2024 elections. The Board found it is not clear whether Meta’s Coordinating Harm and Promoting Crime policy extends to public figures accused of blasphemy in Pakistan or elsewhere.

?? Interesting reads?

• AI Risk Depository: A Comprehensive Database of Risks from AI Systems (MIT)
• Online Gendered Abuse and Disinformation During the 2024 South African Elections (Institute for Strategic Dialogue)
• New Research: Tech Platforms’ Data Access Initiatives Vary Widely (Mozilla)
• Indigenous creators are clashing with YouTube’s and Instagram’s sensitive content bans (Rest of World)

Sign up here to receive updates from the Oversight Board when we announce new events. cases and decisions.

?