About Executive summary Issues Managing risks Stakeholders Transparency and remedy Annex Advancing accessibility through wearable technology In 2025, we launched Meta Ray-Ban Display and Meta Neural Band — a combination of nextgeneration AI glasses and a wristband that uses surface electromyography (sEMG) technology to convert electrical signals created by subtle movements in your muscles into commands allowing hands-free control. Prototypes of the wristband were developed with accessibility in mind and in line with the right to equality and non-discrimination. Based on research involving nearly 200,000 participants, the wristband enables new possibilities for individuals with mobility limitations due to spinal cord injuries, strokes or other disabling events, as well as those who experience tremors or have fewer than five fingers. To protect privacy, the machine learning that translates EMG signals into digital commands is processed locally on-device. Together with AI glasses, our technology empowers people with disabilities to navigate daily life with greater independence. Features such as hands-free calling, real-time translation and live captioning break down communication barriers, while our “call a volunteer” feature — created in partnership with Be My Eyes — allows blind and low-vision individuals to connect with volunteers for real-time descriptions of their surroundings. In building these tools, we engaged directly with disability stakeholders, including the Blinded Veterans Association and U.S. Department of Veterans Affairs’ Blind Rehabilitation Centers, to develop training guides and support accessibility use cases. Continuing responsible AI innovation We released the standalone Meta AI assistant alongside an updated open-source large language model, Llama 4. We also launched Private Processing, a new technology leveraging a trusted execution environment, to enable Meta AI assistant on WhatsApp while preserving privacy protections, including encrypted communications. For details, see our prior due diligence that discusses the human rights benefits of end-to-end encryption. To mitigate risks of misuse and potentially harmful outputs from our AI products and models, the development of Llama 4 included pre-training data filtering, post-training mitigations and system-level mitigations. These are outlined in our developer use guide on AI protections. 2025 Meta Human Rights Report 19