5. How we identify and address salient human rights risks (cont.) (cont.) consent-driven, publicly available dataset that enables researchers to better evaluate how fair and robust certain types of AI models are, with the goal of making them more inclusive. Some languages and dialects may be more We use a combination of AI and human challenging to moderate than others content moderators to help remove policy-violating content on our apps. We use human and machine-based translation to enhance content moderation. We are constantly evolving our reporting flows and integrity classifiers in the languages we support. Hate speech We removed approximately 50 million pieces of hate speech content from Facebook in 2022, 91% of which we proactively found and took action on. We removed 16 million pieces of hate speech content from Instagram, 92% of which we proactively found and took action on. Hate speech made up less than 0.02% of viewed content on Facebook and Instagram. For more on how Meta’s Civil Rights team addresses equality and non-discrimination, see this progress report and this update. Accessibility As technology advances, it’s important that people with disabilities are not left behind. We strive to continually improve our efforts to ensure accessibility across our services and apps. Making these accessible to people with disabilities is a key commitment in our Corporate Human Rights Policy, where we cite the Convention on the Rights of Persons with Disabilities as a core international standard. Our actions: Expanding cross-industry efforts to improve accessibility Virtual, augmented, and mixed reality technologies offer particular promise for people with disabilities, whether through assistive technologies, rehabilitation Human rights report of the 50 million pieces of hate speech content removed from Facebook in 2022 was proactively found and acted on 20

Select target paragraph3