In the face of generative AI, it has become increasingly hard to know if you can trust what you see online. When it comes to News content, eye witness photos and footage are critical for reporting. But often this content surfaces on social media, where there is a high risk of mis and disinformation. Audiences deserve to know how journalists have found and fact checked the content published in stories.

That’s why we developed Content Credentials, an audience facing feature for BBC News that uses the C2PA media provenance standard to encode and share the manual verification check data gathered by BBC Verify journalists. This means audiences can see how the journalists know that it is real and relevant to the story. BBC R&D have been critical to building the C2PA standard which is being adopted across the technology and media industry, and this feature was one of the first examples of a major broadcaster using it in live content.

Audience feedback has shown that the inclusion of the content credentials increased trust in the media by up to 90%.

BBC World Service publish News in over 40 languages. We developed, trialed and scaled BBC’s first generative AI platform to supports journalists to find and adapt key content for global audiences. The project required rigorous evaluation of different models, oversight mechanisms to ensure editorial control and accountability, transparency measures for audiences, and demonstration of value for users and audiences.

There is a critical need to develop new, early interventions for depression, anxiety and psychosis. But to do so will require a huge amount of multimodal, longitudinal data across large, global cohorts. This kind of data only exists in siloed fragments, and to collect new data will require novel approaches to handle the complexity required in a way that is equitable.

In order to show what kind of approaches would be most effected, I worked with Wellcome and an intentional team of technologists, metal health researchers, and lived experience advisors, to develop MindKind, a data collection platform and used it for a randomised control trial with young people in India, South Africa and UK.

The findings informed a follow up phase of strategic work developing the Data Futures report surveying and identifying opportunities to unlock new forms of computational mental health research through investment in data infrastructures.