FREEDOM ON THE NET 2023 The Repressive Power of Artificial Intelligence Ultimately, strengthening information integrity is a long-term challenge that requires long-term solutions. A whole-of-society approach to fostering a diverse and reliable information space entails supporting independent online media and empowering ordinary people with the tools they need to identify false or misleading information. Civic education initiatives and digital literacy training can help people navigate complex media environments. Governments should also allocate funding to develop detection tools for AI-generated content, which will only become more important as these tools grow more sophisticated and more widely used. Finally, democracies should scale up efforts to support independent online media through financial assistance and innovative financing models, technical support, and professional development support. Companies The private sector has a responsibility to ensure their products contribute to, rather than undermine, a diverse and reliable information space. Companies should invest in staff working on issues related to public policy, integrity, trust and safety, and human rights, including regional and country specialists. These teams should collaborate closely with civil society groups around the world to understand the local impact of their products. Without such expertise, companies are ill-equipped to address the myriad of human rights violations and challenges to information integrity that can emerge online and have offline consequences. Companies should develop effective methods to label AI-generated content, which entails using a cryptographic signature, and coordinate with civil society to standardize how the industry documents the provenance of specific content. Companies should also invest in developing software to detect AI-generated content. Companies should ensure transparency and fairness in their policies and decisions, including by being open about how machine learning is used to train automated systems tasked with classifying, recommending, and prioritizing content for human review. They should also refrain from relying on automated systems to remove content without the opportunity for meaningful human review, and establish mechanisms for explanation, redress, and appeal. Finally, companies should work closely with independent researchers who can study the effects their services have on information integrity and free expression. 3. COMBAT DISPROPORTIONATE GOVERNMENT SURVEILLANCE Governments worldwide have passed increasingly disproportionate surveillance laws, and can access a booming commercial market for surveillance tools, giving them the capacity to flout the rule of law and monitor the private communications of individuals inside and beyond their borders. The lack of data privacy safeguards in the United States and around the world exacerbates the harms of this excessive government surveillance. Governments Government surveillance programs should adhere to the International Principles on the Application of Human Rights to Communications Surveillance, a framework agreed upon by a broad consortium of civil society groups, industry leaders, and scholars. The principles, which state that all communications surveillance must be legal, necessary, and proportionate, should also be applied to AI-driven and biometric surveillance technologies, targeted surveillance tools like commercial spyware and extraction software, and open-source intelligence methods such as social media monitoring. In the United States, lawmakers should reform or repeal existing surveillance laws and practices, including Section 702 of the Foreign Intelligence Surveillance Act and Executive Order 12333, to better align them with these standards. Broad powers under Section 702 and Executive Order 12333 have allowed US government agencies to collect and access Americans’ personal data without meaningful transparency or oversight. The US Congress should also close a legal loophole that allows US 32 @freedomonthenet #FreedomOnTheNet