Generative AI Newsletter: Week 51

This week, OpenAI made waves by unveiling o3, an advanced AI model designed to enhance reasoning and promising revolutionary interactions by 2025. Google invites users to test its video generator Veo 2. OpenAI faces a €15 million fine in Italy for data violations, while the French Senate examines the impacts of generative AI on law. Midjourney innovates with Moodboards to customize visual models, and ChatGPT opens up to everyone for real-time searches.

December 22

• OpenAI introduces o3, an AI with enhanced reasoning: OpenAI unveiled o3, an AI model designed to improve ChatGPT's reasoning. It breaks down problems into steps to provide more accurate answers, excels in complex tests, and surpasses coding experts. Currently in testing, o3 will be rolled out to the public in 2025 for enriched interactions.
• Test Google Veo 2 in France starting today: Google recently launched Veo 2, an upgraded version of its AI-assisted video generator, offering 4K renders and smoother human movements. Currently, access to Veo 2 is limited to the United States via Google Labs VideoFX, with a waitlist sign-up. For users in France wishing to try Veo 2, it's possible to bypass this restriction by using a VPN to simulate a connection from the United States. Once the VPN is activated, simply visit the VideoFX site, log in with a Google account, and join the waitlist.

December 20

• Fine for OpenAI in Italy: The Italian data protection authority (GPDP) imposed a €15 million fine on OpenAI, creator of ChatGPT, for illegal use of personal data. The investigation revealed that OpenAI processed user data without adequate legal basis, violating transparency principles and information obligations. Additionally, the company failed to implement an effective age verification system to prevent those under 13 from accessing inappropriate AI-generated content. In response, OpenAI called the sanction disproportionate and announced its intention to appeal. Furthermore, the GPDP ordered OpenAI to conduct a six-month public awareness campaign in Italy to inform about ChatGPT's data collection practices. This decision comes amid increased scrutiny of AI technologies by European and American regulators.
• The Senate analyzes the impact of generative AI on law: Senators Christophe-André Frassa (LR) and Marie-Pierre de La Gontrie (PS) examined the impact of generative AI on legal professions. Their report, presented to the law commission, highlights that this technology, although still emerging, promises major upheavals. They recommend proactive action to integrate generative AI, rather than endure it, to leverage its opportunities while managing its risks.

December 19

• AI models and GDPR, EDPB issues its opinion for responsible AI: The European Data Protection Board (EDPB), which brings together European CNILs, adopted an opinion on the processing of personal data for the development and deployment of AI models, emphasizing the importance of responsible AI compliant with GDPR.

December 18

• Midjourney introduces Moodboards to customize your models: On December 17, 2024, Midjourney launched a new feature called Moodboards, allowing users to customize their generative models by uploading up to 30 images reflecting a specific style or mood. This innovation aims to facilitate the creation of images that match users' aesthetic preferences, simplifying the model customization process.

December 17

• ChatGPT for everyone: OpenAI announced that ChatGPT now allows all its users to access real-time online information, thanks to the integration of internet search. This feature, initially reserved for paying subscribers, is now available to all, offering more accurate and updated responses. This evolution aims to enhance ChatGPT's utility by providing up-to-date information, complementing the data on which the model was initially trained.