Brief

OpenAI has formed a strategic partnership with the Italian media group GEDI to integrate content from its news outlets into OpenAI's products. This collaboration will provide ChatGPT users with attributed quotes, content, and links to GEDI publications like La Repubblica and La Stampa. The aim is to enhance the accuracy and relevance of OpenAI's offerings for Italian users while also expanding GEDI's international reach through advanced translation capabilities. AI

OpenAI has identified and banned accounts linked to an Iranian influence operation that utilized ChatGPT to generate content. The operation focused on topics including the U.S. presidential campaign and the conflict in Gaza, publishing articles on fake news sites and posting comments on social media. Despite these efforts, the operation did not achieve significant audience engagement, with most posts receiving minimal interaction. OpenAI shared threat intelligence with government and industry stakeholders to help disrupt such activities. AI

Banque des Territoires, Polyconseil, and Hugging Face have collaborated to develop a sovereign data solution for a significant French environmental program. This initiative aims to enhance the program's capabilities by leveraging a secure and independently managed data infrastructure. The project underscores a commitment to data sovereignty within critical national initiatives. AI

Cruise, General Motors' self-driving car company, has settled with a pedestrian who was dragged by one of its autonomous vehicles in San Francisco. The settlement is reportedly between $8 million and $12 million. This incident, which occurred last October, led to regulators halting Cruise's operations in San Francisco and a subsequent nationwide suspension of its robo-taxi fleet. Cruise is now beginning to reintroduce its vehicles to the road, starting with testing in Arizona. AI

OpenAI has announced new global partnerships with prominent news organizations Le Monde and Prisa Media. These collaborations aim to integrate French and Spanish news content into ChatGPT, allowing users to interact with summaries and linked articles. The partnerships also involve using the publishers' content to train OpenAI's models, supporting journalism while exploring new revenue streams and audience engagement methods. AI

OpenAI has detailed its strategy for addressing the use of its AI tools in the context of global elections in 2024. The company implemented safeguards to prevent the generation of deepfakes, particularly of political figures, and to avoid expressing political preferences. OpenAI also focused on directing users to authoritative sources for voting information and election results, with ChatGPT providing links to resources like CanIVote.org and encouraging users to consult news outlets such as the Associated Press and Reuters. AI

OpenAI has launched the OpenAI Academy for News Organizations, a new learning hub aimed at journalists and publishers. This initiative, developed in partnership with the American Journalism Project and The Lenfest Institute, offers training, resources, and guidance on the responsible use of AI in newsrooms. The academy focuses on practical applications such as research, translation, and data analysis, while also addressing concerns about accuracy and trust. OpenAI also reiterated its stance on AI training data, asserting that it constitutes fair use and is essential for innovation, while also offering an opt-out mechanism for publishers. AI

Hugging Face has updated its content guidelines and policies to address the responsible use of AI models hosted on its platform. The new policies aim to balance open-source principles with safety considerations, particularly concerning the generation of harmful or illegal content. These updates reflect an ongoing effort within the AI community to establish clearer standards for model deployment and usage. AI

Hugging Face has been chosen by France's data protection authority, CNIL, to participate in an enhanced support program. This initiative aims to assist AI companies in complying with data protection regulations. The program will provide Hugging Face with expert guidance and resources to navigate complex privacy laws. AI

Replit is implementing outbound data transfer limits starting April 7th to combat abuse and manage costs, aiming to improve service quality for all users. Free tier developers will receive 10 GiB per month, with higher limits for paid plans, and additional data can be purchased. The company has adjusted these limits based on user feedback, increasing the allowances for all tiers. AI

A lawsuit has been filed against GitHub, Microsoft, and OpenAI concerning their AI coding assistant, Copilot, with allegations that it may violate open-source licenses. Additionally, Meta AI's Galactica model has drawn criticism for generating problematic scientific content. These controversies highlight the risks and ethical considerations surrounding current AI technologies. AI

AI is increasingly being used in creative fields, sparking controversy among artists as AI-generated works win competitions. This trend raises questions about licensing and the broader impact of automation on creative professions. Additionally, AI is being employed by governments for practical applications, such as identifying undeclared swimming pools in France for tax purposes. AI

Replit has implemented a new policy blocking all internet access to adult content and malware within its computing environment. This change utilizes Cloudflare's family-friendly DNS resolver by default for all user repls. The platform aims to protect its significant user base of teenagers while still providing powerful computing capabilities, and ensures compliance for schools using Replit. AI

Replit has joined the Google for Education Integrated Solutions Initiative, leveraging Google Cloud infrastructure to enhance its coding platform for students and educators. This partnership enables seamless integration with Google Classroom, allowing teachers to assign Replit projects and utilize autograding tools. The collaboration also ensures student privacy with FERPA/COPPA compliance and makes Replit accessible on Chromebooks and other devices, aiming to provide equitable coding education. AI

AI practitioners can engage with governments on AI for good projects, as discussed at the Government & Public Sector R Conference. Panelists shared insights on accessing government data, ethical data usage, and privacy considerations. The R Conference is scheduled to return in Fall 2021. AI

Replit has updated its community standards to foster a safer and more supportive environment for its users. The new guidelines emphasize kindness, inclusivity, and mutual support among coders of all skill levels. Prohibited content includes bullying, hate speech, inappropriate material, spam, illegal activities, and resource misuse, with violations leading to warnings or bans. AI

Replit is phasing out its anonymous coding experience due to Google's new pricing for reCAPTCHA. The platform, which hosts 150,000 concurrent containers and previously relied on Google's free reCAPTCHA to combat abuse, now faces costs exceeding its revenue. To maintain security and service availability, Replit will require users to create accounts to save work and access full features, while offering a limited anonymous experience for trying out code snippets. AI

This podcast episode discusses the nature and dangers of deepfake videos, highlighting their use as a tool for disinformation and state-sponsored information warfare. It explores methods for detecting and resisting these manipulated media, emphasizing the real-world threat they pose in an era of political distrust. The discussion touches upon various technical papers and tools related to deepfake generation and detection, as well as legislative efforts to combat their malicious use. AI