GenAI Newsletter | OpenAI DeepResearch, Mistral Le Chat, AI Action Summit, and more...

Dear readers,?

Welcome to the first edition of our Gen AI Newsletter, where we help you stay ahead of the curve with a curated analysis of the latest Gen AI updates.??

After the earthquake caused by DeepSeek last month, the past few weeks have been nothing short of a rollercoaster. We're witnessing a surge in innovation, from OpenAI's "Deep Research" and Mistral's "le Chat" to major investments in European AI. Best of all: the AI Action Summit in Paris, in which we took part and presented our brand-new study on the Future of Work with AI.?

From new model releases and market insights to ethical debates and practical business applications, we’ve got you covered. You don't want to miss out!

GenAI News Podcast

Don't have time to read this newsletter? Listen to the 14' podcast

Generative AI: Top Highlights at a Glance

Following latest DeepSeek updates, OpenAI has unveiled "Deep Research", a new AI agent integrated into ChatGPT that generates comprehensive, cited reports in 5 to 30 minutes. Powered by its advanced o3 reasoning model (with a slimmed-down version, o3-mini), the tool scours hundreds of online sources including text, images, and PDFs. Aimed at professionals in finance, science, and engineering, Deep Research is available exclusively in the US for Pro tier users at $200/month with a cap of 100 queries monthly.

Mistral introduced le Chat, an AI assistant competing with ChatGPT?- now available on iOS and Android, with enterprise deployment coming soon. Key features include "Flash Answers" delivering up to ~1000 words per second, powered by high-performance Mistral models. The platform offers robust document and image processing (via leading OCR and vision tech), a built-in code interpreter for sandboxed analysis and simulations, and state-of-the-art image generation using Black Forest Labs Flux Ultra. Pricing starts with a free tier,Pro tier available at $14.99/month, and additional Team and Enterprise tiers.?

Following the passing of Alain Dorval, the iconic French voice of Sylvester Stallone, ElevenLabs utilized AI to recreate his distinctive voice for a recent French-dubbed film. This development has sparked ethical debates regarding the use of AI to replicate deceased actors' voices, with concerns about potential job displacement in the voice acting industry. Proponents on the other hand argue that such technology can preserve beloved performances and maintain continuity in film franchises.

The AI Action Summit, held in Paris on February 10-11, 2025 and co-chaired by Emmanuel Macron and Indian Prime Minister Narendra Modi, convened global leaders, tech executives, and researchers to discuss the future of artificial intelligence. Key developments from the summit include:

• French President Emmanuel Macron announced a €109 billion investment in AI, emphasizing France's clean electric power as an advantage for managing AI's high energy demands. He urged European cooperation and a unified strategy to compete with the U.S. and China in AI development.?
• A group of 20 major corporations, firms, and startups rallied behind a plan to invest €150 billion in European AI over the next five years. The initiative, led by venture capitalist firm General Catalyst, wants to bridge the gap between investors and startups to "unlock Europe’s full potential in AI".
• A new partnership was announced at the summit on Monday between French start-up Mistral AI and Helsing, a European defense technology company, who will work together on AI systems in defense.
• The United States and the United Kingdom declined to sign a declaration endorsed by approximately 60 countries, which aimed to ensure AI technology remains "safe, secure, and trustworthy." Concerns over potential stifling of innovation due to excessive regulation were cited as primary reasons for their refusal.

Breaking news

1. DXC helps European Space Agency launch GenAI Agents
2. Legal assistant Harvey raises $300M Series D led by Sequoia
3. Meet Adobe's Firefly Video Model: AI-Powered creation with unparalleled creative control

Insights & New Models

> Business News & Market Insights

#1. DeepSeek’s AI technology has been banned by several countries and government agencies due to privacy and security concerns. Italy banned DeepSeek over GDPR violations, while Taiwan cited security risks. In the U.S., the Navy, Pentagon, NASA, Texas, and Congress have restricted its use, amid concerns over data leaks to China.

#2. France Travail and Mistral AI are launching AI tools to assist job seekers and counselors. ChatFT helps with writing, while MatchFT contacts candidates via SMS.. This initiative boosts Mistral AI’s position in Europe, following a €600M funding round and partnerships with the French Ministry of Armed Forces and AFP.

#3. In the Latest LLM Race, Google leads with an average top Elo score of 1382 and 5.5 models released per quarter in 2024, showing a sharp acceleration since early this year. OpenAI follows with a 1365 Elo and 3 models/quarter. DeepSeek is close behind at 1358 Elo with 1.8 models per quarter—potentially underestimated—as its release pace is set to accelerate later in the year. xAI with the slowest pace at 0.5 models/quarter and a 1288 Elo, Anthropic at 1284 Elo and 1.8 models/quarter, Alibaba at 1282 Elo and an aggressive 3.5 models/quarter, Meta at 1262 Elo and 2.5 models/quarter, and Mistral at 1252 Elo with 2.0 models/quarter.

#4. I'M A PROMPT redefines the image by harnessing artificial intelligence to push back the boundaries of visual creation. Each image, advert or poster in the magazine explores the creative potential of technology while asserting a strong and amplified artistic direction. Take a look at this living portfolio and artistic manifesto !

#5. The U.S. Copyright Office has issued new guidance clarifying that works generated solely by AI without human involvement are not eligible for copyright protection. However, if a human artist significantly modifies AI-generated content, incorporating personal creative elements, the resulting work may qualify for copyright.?

#6. OpenAI co-founder John Schulman has departed from AI startup Anthropic five months into his tenure to join former OpenAI CTO Mira Murati's new, yet-to-be-named venture. Schulman initially left OpenAI in August 2024 to focus on AI alignment research at Anthropic.

#7. Artefact and Odoxa released a study on The Future of Work with AI” based on a survey of over 1000+ French workers and 30+ interviews.

Key findings are:

• 12% of French workers use AI at work and individual adoption is unstoppable.
• AI users at work save 57 minutes per day and there is a latent collective productivity boom ready to be unlocked with AI agents.
• There are two main types of use-cases with high impact on work: Knowledge assistance & Workflow automation.
• Knowledge and expertise management will be transformed by AI assistants, for more flexible careers and faster junior ramp-up.
• Like in any automation revolution, work is replaced with supervision and higher-value activities.
• AI shifts focus back to humans. The study shows that 33% of AI users at work exchange more with their colleagues thanks to the time they save and 58% find their work more collaborative. Back-office workers replaced with AI are often reallocated to client-facing tasks.
• AI Agents are best experimented and deployed by small teams saturated with tedious & repetitive tasks.
• To foster experimentation, enterprises need to empower their teams with methodological support and Data& AI tools. Workers need to feel safe to experiment.

#8. Elon Musk, co-founder of OpenAI, has led a consortium in an unsolicited $97.4 billion bid to acquire the nonprofit entity overseeing OpenAI. This move intensifies the ongoing rivalry between Musk and OpenAI CEO Sam Altman, who have previously clashed over the organization's strategic direction after Musk departed OpenAI in 2019 to start his own AI company xAI. Musk's proposal emphasizes a return to OpenAI's original mission of open-source, safety-focused AI development. The bid has been met with resistance, with Altman and others expressing concerns about Musk's intentions and the potential implications for OpenAI's future and capacity for innovation.

#9. A recent study by Microsoft and Carnegie Mellon University reveals that increased reliance on Gen AI tools in the workplace may lead to reduced critical thinking among employees. The research involved 319 knowledge workers who provided 936 examples of AI usage in their jobs, indicating that higher confidence in AI outputs corresponded with diminished engagement in critical evaluation. The study warns that this dependence could result in the "deterioration of cognitive faculties" necessary for effective problem-solving. To mitigate these effects, the researchers suggest designing AI tools that encourage user engagement and critical assessment, such as providing explanations of AI reasoning and areas for user refinement.

#10. Despite widespread acknowledgment of AI’s efficiency benefits, a University of Lausanne study found that while managers saved an average of 3 hours per week using AI, 36% of them admitted to wasting more than half of this time. Most users simply revert to more of the same tasks rather than using the freed-up hours for meaningful activities like learning, connecting with others, or improving well-being. To optimize these time savings, experts recommend that employees track their time gains, develop strategic plans for reallocating it, and receive ongoing guidance from managers to ensure that the efficiency gains lead to meaningful productivity improvements and personal development.

#11. French operators have signed exclusive partnerships with GenAI actors. Free Mobile subscribers can now enjoy 12 months of Mistral AI’s Chat Pro, offering unlimited messages and advanced features. Bouygues Telecom gives all its customers free access to Perplexity Pro. Meanwhile, Orange teams up with Mistral AI to integrate AI-driven solutions like Chat Pro and Codestral, aiming to optimize networks and boost enterprise productivity.

> New Models

#1. Mistral AI has launched Mistral Small 3, a 24B-parameter model optimized for speed and efficiency. With 81% MMLU accuracy and 150 tokens/s, it outperforms larger models like Llama 3.3 70B while running 3x faster. Ideal for low-latency AI tasks, it excels in function calling, fine-tuning, and local inference.?

#2. Google has announced significant updates to its Gemini 2.0 AI model lineup, introducing the cost-effective Gemini 2.0 Flash-Lite, which offers improved performance over previous models while maintaining speed and affordability. The company also released an experimental version of Gemini 2.0 Pro, designed to handle complex tasks such as coding and advanced problem-solving, featuring a context window of 2 million tokens and the ability to integrate external tools like Google Search and code execution.

#3. OpenAI is set to launch GPT-5 soon, marking a significant milestone in its AI model development. This new model will integrate various technologies, including the reasoning capabilities introduced with o1, to provide more reliable and consistent responses. GPT-5 aims to simplify the user experience by eliminating the need to manually select models. This advancement is crucial for applications like customer service and research that rely on accurate interactions. By unifying its tools, OpenAI seeks to make AI more intuitive and effective across a wide range of tasks.

Focus of the Week

?? DeepSeek R1: Proving That “Bigger is Not Better” (But Not Without Controversy)

DeepSeek R1 is shaking up the AI market with an ultra-optimized approach: GPT-4-o-level performance, but with inference costs slashed by 30x and training expenses at just $6M. Through key innovations (MoE, GRPO, optimized RLHF), DeepSeek challenges the idea that raw power is the only path to AI dominance. But this revolution comes with controversy: allegations of training on OpenAI data, concerns over censorship, and GDPR uncertainties around its API. DeepSeek R1 is both a technological breakthrough and a regulatory grey zone, pushing the limits of what’s possible in cost-effective, accessible AI.

French version here.

Generative AI Software & Features

#1. Apple Intelligence will expand to support more languages, including French, German, and Spanish, starting in April. CEO Tim Cook announced this during the Q4 2024 earnings call. The AI suite initially launched with U.S. English, adding more localizations in December. A new version of Siri with on-screen context understanding will also roll out soon.

#2. Meta's AI chatbot now features memory, allowing it to recall details from previous chats and offer personalized recommendations based on user data from Facebook and Instagram. This will be available on Facebook, Messenger, and Instagram in the U.S. and Canada. Users can't opt out of the data use. The updates have raised privacy concerns..

#3. Amazon is preparing to launch a significant upgrade to its Alexa voice assistant, integrating Gen AI capabilities to enhance user interactions. The revamped service, internally referred to as "Remarkable Alexa", is designed to handle complex queries, perform multiple tasks from a single prompt, and remember user preferences to offer personalized recommendations. The company partnered with Anthropic to utilize their Claude AI models for improved functionality. This objective is to transform Alexa into a more conversational assistant to convert users into paying customers.

Thoughts of the week by Hanan Ouazan, Managing Partner & Global Lead AI Acceleration

Reinventing Work - The True Impact of Automation and Generative AI

Generative AI is profoundly reshaping the way we work, introducing tools that automate a wide range of tasks - from coding and data analysis to content creation. These tools don’t replace human expertise; they redefine it. Experienced professionals - whether engineers, analysts, or creatives - are becoming indispensable, not for execution, but for supervising, refining, and integrating the outputs produced by these technologies.

A concrete example: industrial automation. Manufacturing goods in factories has become faster and more cost-effective thanks to technology. This didn’t eliminate workers but transformed their roles—shifting many to quality control, performance monitoring, and process management. At the same time, demand for engineers capable of redesigning production systems and envisioning new solutions has skyrocketed.

The same shift is happening across sectors. Automation doesn’t eliminate existing jobs; it evolves them, moving value creation toward two critical areas:

• Supervision and optimization: expertise to monitor and continuously improve automated systems.
• Rethinking processes: talent capable of integrating these technologies into a broader vision that fundamentally transforms how value is created.

However, simply layering automation onto existing processes isn’t enough. It risks overloading systems without making them meaningfully more efficient. The real challenge lies in rebuilding these processes from the ground up, leveraging these tools to redefine workflows and unlock entirely new productivity drivers. Without this foundational shift, the gains will remain marginal, and true transformation will remain out of reach. *

Automation doesn’t eliminate jobs - it shifts where value is created. Tools alone won’t transform organizations; it’s how they’re embedded into a reconstruction of the fundamentals that matters. Those who successfully evolve their processes and talent around these technologies won’t just gain a competitive edge - they’ll redefine the rules of the game.