When AI becomes teacher, therapist, and boss

In partnership with

👋 Dear Dancing Queens and Super Troupers,

It’s back-to-school season… and artificial intelligence has decided to play the teacher, the therapist, and even the multinational CEO.

The week begins with a bang as Meta is forced to add guardrails to its chatbots after bizarre interactions with teenagers.

Mark Zuckerberg’s company, already accused of ignoring addiction and harassment, this time had to retrain its AI so it would stop giving minors advice on suicide or eating disorders.
A “parameter error,” according to them. That’s one way to put it lightly.

Meanwhile, Beijing isn’t fussing around: China has unveiled its 2035 roadmap to become an “intelligent society,” with AI embedded in 90% of daily uses, from environmental monitoring to governance.

The stated goal: reshape the global economy and position AI as a “global public good.” Translation: they want the whole world plugged into their stack. The ambition is planetary, and the timing comes just as Western voices warn of a bubble.

In the West, Nvidia is setting the pace. Jensen Huang, its rockstar CEO, just announced a record quarter at $46.7 billion, and promises AI will bring us the four-day workweek.

Well, “promises”… while clarifying that we’ll be “busier than ever.” In other words: fewer days in the office, but more projects, more productivity, and more $4-trillion data centers to feed its chips.

On the darker side, ESET has uncovered PromptLock, an experimental ransomware that literally codes itself using AI.
Each attack is unique, undetectable, and potentially within reach of a lone individual. For now, it’s only in the lab, but it illustrates a new era where cybercriminals no longer need entire teams of developers. AI becomes their weapon of choice…

And in France? The Ministry of Education has announced a sovereign AI for teachers. An assistant meant to “clear the brush” in lesson prep and analyze students’ difficulties.
Budget: €20 million under France 2030, launch expected in 2026. But one question remains: who was really waiting for an official AI when ChatGPT, Gemini, or Mistral are already being used in teachers’ lounges?

From Meta playing arsonist-firefighter to Nvidia redrawing our workweeks, from China trying to reshape the planet to hackers outsourcing their scripts to AI, this back-to-school season proves one thing: AI isn’t just changing our tools — it’s redrawing our lives, our institutions, and our fears. So, ready for the new term?

Here’s this week’s lineup:

👉️ Meta AI: from unhealthy flirting to late guardrails 🥵​​​​​

👉️ Plan 2035: China bets on AI to reshape the world 🤦‍♀️​

👉️ Nvidia: AI promises the four-day week… with more work 👽

👉️ PromptLock: the ransomware that writes its own attacks 🧐​

👉️ When AI lands in the classroom 😎​​

If we've forwarded this letter to you, subscribe by clicking on this link !

Keep This Stock Ticker on Your Watchlist

They’re a private company, but Pacaso just reserved the Nasdaq ticker “$PCSO.”

No surprise the same firms that backed Uber, eBay, and Venmo already invested in Pacaso. What is unique is Pacaso is giving the same opportunity to everyday investors. And 10,000+ people have already joined them.

Created a former Zillow exec who sold his first venture for $120M, Pacaso brings co-ownership to the $1.3T vacation home industry.

They’ve generated $1B+ worth of luxury home transactions across 2,000+ owners. That’s good for more than $110M in gross profit since inception, including 41% YoY growth last year alone.

And you can join them today for just $2.90/share. But don’t wait too long. Invest in Pacaso before the opportunity ends September 18.

Invest While You Still Can

_{Paid advertisement for Pacaso’s Regulation A offering. Read the offering circular at invest.pacaso.com. Reserving a ticker symbol is not a guarantee that the company will go public. Listing on the NASDAQ is subject to approvals.}

🔥 If you have 15 minutes

1️⃣ Meta AI: from unhealthy flirting to belated guardrails

The summary: Meta tightens the rules. Its AIs will no longer discuss self-harm, eating disorders, or suicide with teenagers. Also gone: romantic conversations or open access to user-created characters. These guardrails follow revelations of dangerous exchanges, including “sensual” dialogues.
Meta is rolling out temporary protections for teenagers in English-speaking countries but already faces a wave of U.S. regulatory investigations.

Details :

• Teens and AI: off-limits : Meta redraws the boundaries: no more minors getting advice on self-destruction or flirting with a chatbot. The company now redirects to specialized resources rather than risking dangerous drift.

• Filtered characters : User-created AI profiles remain off-limits for young people until Meta installs lasting safeguards. Access will be restricted to a small circle of “safe” characters.

• Embarrassing revelations : Reuters sparked the fire by exposing a document where AI was allowed to have “sensual” conversations with minors. Meta quickly labeled the content “erroneous” and removed it.

• Political and legal pressure : Senator Josh Hawley is preparing a congressional inquiry, while Texas Attorney General Ken Paxton wants to investigate whether Meta misled children about the mental health advice given by its bots.

Why it's important: Meta is walking a tightrope: attracting young users while keeping them safe. By locking down its AIs, the company is trying to prevent another scandal and prove it takes child safety seriously. But with safeguards still temporary, one question looms: is this just a quick patch — or a genuine rebuild of trust?

2️⃣ Plan 2035: China bets on AI to reshape the world

The summary: China has just announced its ten-year plan: by 2035, artificial intelligence must permeate its entire economy and transform society. The State Council wants to make it a central engine of growth, with 90% adoption by 2030.
Six pillars are targeted: science, welfare, industry, consumption, governance, and international relations. Beijing dreams of an “intelligent society” while presenting AI as a global public good. But between planetary ambitions and models still lagging behind the U.S., the gamble remains risky.

Details :

• A roadmap through 2035: China sets a precise agenda: generalized AI by 2027, 90% adoption in 2030, and a complete economic overhaul by 2035. A carefully scripted timeline.

• Six domains under digital steroids : From environmental monitoring to social sciences, AI must strengthen research, reinforce science, stimulate industry, smooth consumption, and soften diplomacy. A digital graft into every vital organ of the country.

• Algorithmic soft power : The State Council stresses that AI should become a “global public good.” Beijing promises open source, cooperation with the Global South, and support for the UN — a way of projecting influence beyond the Great Wall.

• The gap with Washington is narrowing : In 2023, U.S. models outperformed Chinese AI in reasoning tests by 13%. A year later, the gap is down to 8.1%. Beijing is doubling down on data, security, and energy to close the distance.

Why it's important: If this plan succeeds, China could shift from “factory of the world” to “brain of the planet.” But between software that can still be temperamental and XXL ambitions, the country is playing a chess game where every move will be watched closely by Washington, Brussels, and the rest of the globe.

3️⃣ Nvidia: AI promises the four-day week… with more work

The summary: Jensen Huang, Nvidia’s CEO, predicts that AI will “transform every job” and could make the four-day week as natural as the disappearance of seven-day workweeks in the 19th century.
But don’t expect a slowdown: AI accelerates ideas, boosts productivity, and rewrites the rules of the game. With record profits ($46.7 billion last quarter), geopolitical ambitions, and plans for multi-trillion-dollar AI factories, Nvidia is positioning itself as the architect of this revolution.

Details :

• AI as a turbo for social change : For Huang, every industrial revolution reshapes habits. The next one? Four-day weeks, more free time, but productivity skyrocketing.

• Endless creativity : AI cuts heavy tasks, frees up time, and fuels leadership creativity. The more machines advance, the more ideas multiply — a nearly virtuous circle.

• Galactic profits : With $46.7 billion earned in three months and a market cap above $4 trillion, Nvidia dominates the market, powered by its Blackwell Ultra chips.

• Planetary AI factories : Huang announces $3 to $4 trillion worth of infrastructure by the end of the decade, spanning robotics to data centers.

• The four-day week already tested : Trials in the UK and North America show +24% productivity, fewer burnouts, and improved quality of life. In the Netherlands, 32 hours of work is already enough to validate the model.

Why it's important: Because this “Nvidia week” could well become the norm. AI isn’t just removing tasks — it’s reprogramming work. Fewer hours, greater efficiency, and perhaps, for the first time in a long while, an economy that breathes as much as its workers.

4️⃣ PromptLock: the ransomware that writes its own attacks

The summary: A new ransomware is redefining cybercrime. PromptLock, freshly detected by ESET Research, uses AI to invent its own attacks on the fly.
It generates unique, real-time attacks that are nearly undetectable by traditional antivirus software. For now, it’s mostly a lab toy, but it points to a future where a single hacker could shake the cyber world.

Details :

• Custom-made scripts : Using gpt-oss-20b and the Ollama API, PromptLock creates Lua scripts on the spot. Windows or Linux, it doesn’t matter — the malware adapts its attack plan to each machine.

• Easy hacks, guaranteed chaos : No need to be a coding genius anymore. AI replaces armies of developers, making complex attacks accessible to almost anyone.

• Antivirus confused : Each execution generates a different script. Classic defenses scratch their heads and struggle to detect the threat.

• Still in testing, not in the wild : Don’t panic yet: PromptLock hasn’t struck outside labs. But its potential already worries cybersecurity professionals.

Why it's important: PromptLock is proof that AI is leaving the office for the dark side of the web. It turns a lone hacker into a small army, able to create unique attacks with every click. Digital security will have to adapt — quickly and smartly — or risk being outpaced by a malicious, cunning intelligence.

5️⃣ When AI lands in the classroom

The summary: Just days before the new school year, Élisabeth Borne confirmed the rollout of an AI tool for teachers. This assistant promises to help them prepare lessons and track student progress. An ethics charter will frame its deployment. But in a context where teachers already use ChatGPT, Gemini, or Mistral, one question arises: isn’t the state chasing a train that has already left the station

Details :

•  A backup brain for the classroom : The AI will act as a teaching assistant: quick resource searches, time savings estimated at 57 minutes per day, and personalized tracking to spot student difficulties and progress.

• The minister tempers enthusiasm : Élisabeth Borne insists the AI is just an “auxiliary brain.” No question of replacing critical thinking or the central role of teachers. The tool is support, not a pilot.

• Consultation and guardrails : Ethics charter, consultations with teachers, parents, students, and unions: the ministry wants to set boundaries to ensure responsible use, where technology supports but doesn’t supplant.

• A sovereign €20 million project : Funded under the France 2030 plan, the state wants a “made in France” AI, scheduled for 2026–2027, to retain control over the technology.

• Timing in question : With ChatGPT or Gemini already in classrooms, the initiative risks looking like a patriotic gadget out of step with reality

Why it's important: This initiative highlights France’s dilemma: building a sovereign alternative to AI while staying credible against foreign giants. Beyond the tool itself, it’s a question of training and trust in teachers, who are already autonomous and connected.

❤️ Tool of the Week: nano-banana, Google’s Photoshop for dummies

Integrated into Gemini at the end of August, this free image generator/editor is going viral. Simple and impressive, it lets you transform any photo with just a written prompt.

What’s it for?

• Change appearance : hairstyle, clothes, age, visual style (1920s, Renaissance, etc.).

• Edit backgrounds : teleport yourself to New York or into an impressionist painting.

• Merge multiple photos : add an object or an animal from one picture to another.

• Redecorate a room : insert furniture, a TV, a bookshelf to reinvent your interior.

• Create in 3D : generate an isometric version of an object.

• Realistic retouching: the tool respects faces, avoiding the grotesque distortions seen in rival apps.

• How to use it?

  - On PC: go to gemini.google.com, upload an image, type your request, and click Send.

  - On mobile: download the Google Gemini app (iOS/Android), tap the + button, upload a photo, and enter your prompt.

  - Pro tip: download your generated images right away — max quality sometimes disappears when servers are overloaded.