BSR has previously emphasized the importance of human rights by design in the technology industry. Today, Google announced the launch of a product that took this approach with its new celebrity recognition application program interface (API) for use in the media and entertainment industry. The API will enable Google enterprise customers to identify celebrities in their content at a frame-by-frame or scene-by-scene level using a database of celebrity images licensed by Google.
Google commissioned BSR to conduct a human rights assessment of the API while it was in development to identify, prevent, and mitigate potential human rights impacts. This deliberate integration of human rights into the product design process is best practice—and although it is never possible to eliminate all human rights risks, the result of this collaboration is an API with a significantly lower human rights risk profile.
Working collaboratively, BSR and Google identified a range of potential issues, challenges, and dilemmas relating to the field of celebrity recognition products. These were further analyzed by BSR for potential human rights impacts, such as privacy, freedom of expression, security, child rights, non-discrimination, and access to culture. BSR then developed recommendations to address these impacts, which Google integrated into key elements of the API’s development.
Among the BSR recommendations adopted by Google prior to the API’s launch were the following:
- Applying “Service Specific Terms” to limit the use of the celebrity recognition API to professionally filmed media content (e.g., not user generated content) that the customer owns or has adequate consent to use.
- Restricting inclusion in the celebrity database to individuals whose primary profession involves voluntarily being the subject of public media attention.
- Implementing an opt-out policy to allow celebrities to request removal from the Google-managed celebrity database.
- Establishing a customer “whitelisting” process whereby the customer must be within a qualifying industry (entertainment, media, sports), declare an allowable use case, and agree to only use professionally filmed media.
It is BSR’s view that, taken in combination, these measures serve to prevent, avoid, and mitigate potential adverse impacts and provide Google with a firm basis to reduce risks to human rights.
The result of this collaboration is an API with a significantly lower human rights risk profile.
In addition to actions that can be taken by Google, BSR’s recommendations also included action that can be taken by two other important constituencies: other providers of celebrity recognition tools and the media and entertainment industry.
Google is only one of several companies currently offering a celebrity recognition tool, and it is important that the entire industry deploy approaches that address human rights impacts. For this reason, BSR concluded that industry-wide approaches (such as industry principles or standards), public policy, and regulation may be needed to mitigate the potential human rights risks of these tools.
BSR’s assessment also emphasized the crucial role played by media and entertainment companies using the API responsibly. For this reason, BSR recommended that companies making use of celebrity recognition products and other facial recognition tools should undertake their own human rights due diligence. This builds upon BSR’s perspective that the users of artificial intelligence products need to be much more engaged in understanding their impacts.
Google’s deliberate integration of human rights into the product design and development process—prior to the launch of a product—provides an excellent case study, and we encourage other companies to adopt similar practices as they develop their own products.