👋 Inside this week's edition:

🧠 Paris-based Outline wants to turn financial planning from a monthly spreadsheet exercise into a strategic tool for decision-making, with AI agents that can build forecasts, model scenarios, and answer CFO questions in minutes. Co-founder and CEO Antoine de Mereuil explains the hows and why as the company lands a $3M round to accelerate its development.

Chris O'Brien + Helen O'Reilly-Durand

SPONSORED

For today’s AI-native, data-driven startups, security is no longer optional. Customers expect it, investors demand it, and growth depends on it. 

Learn how leading AI-powered companies—Granola, Clay, Factory, and more—prioritized trust from day one and used Vanta to turn security and compliance into a competitive advantage. From faster SOC 2 audits to six-figure savings and ISO 42001 adoption, see how these startups are building trust in the AI era.

Download it here

Headlines

🗞️ The AI giants are warning that their own technology could destroy humanity. Nvidia’s Jensen Huang and Donald Trump aren’t buying it. After former Anthropic/OpenAI engineer Jacob Coxon quit over fears AI could endanger humanity, Anthropic co-founder Dario Amodei called for a global slowdown in AI development, backed by OpenAI’s Sam Altman and Google DeepMind’s Demis Hassabis. Then, during a live podcast in Los Angeles, Huang took a call from Trump and put the president on speakerphone. Both pushed back against slowing the race, with Huang arguing that the US could remain number one in AI and develop it safely. Behind the safety debate, Les Echos points to a bigger question: with billions at stake, record valuations to justify, and Chinese models rapidly closing the gap, who actually benefits from a slowdown? | Les Echos

🗞️ Europe wants to win the AI race without building the best AI. In her State of the Union speech, Ursula von der Leyen outlined a “third way”: rather than chasing the US and China for ever more advanced models, Europe should focus on deploying AI across industries where it already has strong companies, data and expertise. But with US AI investment running at a vastly different scale and Europe still dependent on foreign chips and critical materials, the strategy raises a bigger question: is this a smart industrial strategy, or a way of reframing Europe’s AI deficit? | Usine Digitale

🗞️ Mistral AI is taking its models into the oilfields. The French AI unicorn has signed a three-year partnership worth more than €100 million with TotalEnergies to develop new frontier AI models for oil and gas exploration, reservoir characterization and engineering. The project will combine Mistral’s AI capabilities with a century of TotalEnergies’ geoscience data, and the two companies will also set up a joint scientific lab. The aim is to generate multiple scenarios - from identifying new exploration opportunities to extending the life of existing projects - and help TotalEnergies’ experts analyze complex data and make decisions faster. | Le Monde, TotalEnergies

🗞️ Women are being left behind in the AI boom, with the gender gap widening at every level of the industry. According to a LinkedIn study covering 27 countries, women account for just 29% of entry-level AI roles globally, compared with 48% across the wider job market. The gap widens with seniority: women hold just 25% of AI management roles and 13% of executive positions, versus 35% and 23% across the wider market. France fares somewhat better, with women making up 18% of AI executive teams. That's five points above the global average. Meanwhile, Paris, Grenoble and other French cities are emerging as AI skills hubs. | Les Echos, LinkedIn Study

🗞️ French companies are racing to deploy AI agents, but few are watching what those agents actually do. A Wavestone study of 30 public and private organizations found that just 17% had controls limiting what AI agents could access or do, while only 15% had adapted identity management for agents. Even more worrying, just 12.5% had a response plan for an AI agent going off the rails, and only 8% of organizations using agentic AI were actively monitoring its security. “Nobody is monitoring whether AI is being attacked or goes off the rails,” warned Wavestone partner Gérôme Billois. | Les Echos, Wavestone

🗞️ Paris is betting on driverless public transport. Île-de-France Mobilités plans to deploy more than 100 autonomous 8–12-seat minibusses by 2030, starting with a first vehicle in 2028 and 20 in 2029. The €20m-a-year program will target low-frequency routes, business parks and public services outside the main transport network, particularly for shift workers at airports, hospitals and Rungis. Trials will begin around Paris-Saclay, Saint-Quentin-en-Yvelines and Versailles Grand Parc, with a European supplier preferred where possible. | L’Usine Digitale

🗣️Responsible AI🗣️

A look at the big stories from recent weeks, by Responsible & Frugal AI expert James Martin, of BetterTech.

Forgive us for piling on to the AI pile-on du jour, but Jacob Coxon’s calling his former bosses irresponsible was under-discussed on two specific angles. 

First, AI leaders have been clueless about their true responsibilities for at least seven years. In 2019, Dario Amodei, then at OpenAI, decided GPT-2 was “too dangerous to release”. He thereby invented myth marketing, a.k.a. exaggerating risk to generate hype, a.k.a. ‘fake it until you make it’. For Yann LeCun, “everyone should make fun of them now”, as then, for this. He has a point.

Secondly, even if AI kills us all *one day*, it is already killing too many people *right now*. 

• Just one planned Virginia data center could cause $53-99 million in health-related economic damages every year, re. EmPower Analytics… and the US is currently doubling its number of data centers
• An increasing number of these facilities are powered by natural gas such as methane, and are already shortening lives, for example near Elon Musk’s Colossus supercomputers in Memphis
• And let’s not even start on the murder sprees and suicides facilitated by generative AI so far (OpenAI rejects responsibility for the latter).

As for the planet, European sovereign AI ‘hero’ Mistral AI just pocketed a crisp €100 million to help TotalEnergies find more petrol. Doubly ironic considering 1/. The French company has to date been the most transparent of all model makers as to its environmental impacts, and 2/. AI is currently boosting oil and gas sector productivity so much that “enabled” emissions could be 3-13 times greater than current data center emissions, according to a widely shared recent study by Will and Holly Alpine.

Another downer for France: its poor results in the international ‘PISA’ tests, which rank 15-year-old pupils worldwide in three key subjects. The culprit was quickly found. “Those who do best in sciences are those who don’t use AI”, said OECD education expert Eric Charbonnier when presenting the study, re. Franceinfo’s report.

Meanwhile, AI continues to make our lives even more effortless…by spying on us.

Apple Watches will soon generate AI-powered recaps of their owners’ days by using their microphones and other sensors, according to Bloomberg (via 9to5 Mac).

And Anthropic is reportedly using AI to monitor anti-AI activists and other threats, no less. In one case, Amodei’s company detected murderous threats against him, reported them to the police and… refused to hand over evidence when asked for it. Could the Anthropic CEO’s next outlandish claim be a Judge Dredd-style “I am the law”?

Join James Martin for “Using AI without costing the Earth: Frugal AI,” a free lightning lesson on Maven, today, September 17, at 3 pm CET. Sign up here! (replay available afterward)

🧠 Outline Wants to Free the CFO From the Spreadsheet

Paris-based Outline has raised $3 million to build an AI-native financial planning platform for CFOs. Founded in June by Antoine de Mereuil and Hubert Jaouen, the startup connects accounting, payroll, billing, and CRM data so finance teams can automate forecasting, run scenarios, and answer questions that still require hours or days of spreadsheet work.

Despite only launching in June, the startup had already signed around 10 finance teams, including those at Zelty, Shares, and HarfangLab, before the funding round was announced. For Outline Co-Founder and CEO Antoine de Mereuil, the opportunity started with a simple observation: Finance teams had become strategically important without getting the tools to match.

“Our analysis was that Finance had been put to one side as a support function,” he said. “But they should be able to play a much more strategic role in the company.”

📆 Events 📆

📆 OpenAI Builder Lounge Paris | September 22 | Join OpenAI for an evening of hands-on building, live demos, and discussion in Paris, co-hosted with Photoroom. Hear how top teams approach building useful agents, see what is new with Codex, and work alongside local founders, CTOs, and engineers. | Request to join

📆 AI Engineer in Paris | September 23-24 | AI Engineer Paris 2026 returns to Station F on September 23–24 for two days of technical talks, practical workshops, and AI engineering research. | Get Tickets

📆 AI Shift #2 @50 Partners | Paris | October 1 | The 50 Partners Tech team is delighted to invite you to the "AI Shift" event at the Loft 50P: a concrete immersion in AI applications.​ The concept: a series of sessions dedicated to analyzing AI-induced upheavals in the technology innovation ecosystem, followed by time to discover new solutions and products through live demos. | Apply

📆 ai-PULSE | Paris | October 13 | Join leaders and engineers for a one-day technical conference dedicated to AI breakthroughs, research, and product demonstrations. | Register

📆 AI Meets Multi-Omics: Enabling Precision Medicine | Paris | October 15 | Institut Curie| Speaker: Nataša Pržulj, Department of Life Science, Barcelona Supercomputing Center & ICREA (Barcelona, Spain) & Department of Computer Science, University College London (London, UK) | Information

👋🏻 If you’re enjoying The French Tech Journal, support the project by forwarding it to friends and sharing it on your social networks. You can also comment on this post. And if you have ideas for stories, tips, or just want to harass us, shoot us an email: chris@frenchtechjournal.com / helen@frenchtechjournal.com 👋🏻‌