J’ai donné accès à toutes mes données à une IA : qu’est-ce qui s’est passé ?

J’ai donné accès à toutes mes données à une IA : qu’est-ce qui s’est passé ?

I gave an AI access to all my data: what happened?

by Silicon Carne · View the channel on PSI

Humanities, Social Sciences & Thought AI & Society UBDigital and Information technology: general topicsUBJDigital and information technologies: social and ethical aspects
🎙 Silicon Carne 👥 99K 📅 September 21, 2026 ⏱ 37 min 👁 29K 📄 debate 🧭 2026-10-16
Available in: English (current) Français

Keywords

AI agentproactive assistantdata accessautonomous decisionuser control

Summary

The video is a panel discussion about AI agents, focusing on three products: Instinct, Muse, and Qapten. Host Carlos Diaz shares his experience giving Instinct full access to his emails, calendar, and messages; after 48 hours, the agent proactively alerted him to a scheduling conflict and a business opportunity, impressing him. Guest Michel Levy-Provençal, founder of Qapten, explains his product’s philosophy of user-controlled capabilities, while Pierre Gaubil questions whether such tools qualify as true agents if they require validation before acting. The conversation covers the ’economy of laziness’—automating tedious tasks like checking subscriptions or bank fees—and highlights an anecdote of someone creating an agent to communicate with his wife. Meta’s launch of Muse is discussed as a major entry by a tech giant, raising concerns about data privacy and the shift from reactive chatbots to proactive, autonomous systems. The panel debates the ethical limits of delegating decisions and the risk of losing cognitive sovereignty. The video contains a sponsored segment of about 84 seconds.

164 words

Critical Evaluation

Value of the Information & Strength of the Argument

The video provides valuable firsthand insights into the capabilities and user experience of a proactive AI agent, distinguishing between push and pull modes of interaction. It also contrasts different product design choices regarding user control. The argumentation is persuasive through concrete examples and expert opinions, though it leans on anecdotal evidence and personal anecdotes rather than systematic evaluation. The debate among the three participants enriches the discussion, offering varied perspectives on autonomy and safety.

Scientific Rigor, Source Quality, Title Accuracy

The video does not reference formal scientific sources, instead relying on product launches, company announcements, and personal experiences. The rigor is limited by the promotional context (e.g., featuring Qapten’s founder) and the lack of independent verification. The title accurately reflects the content, as the host’s experiment is a central element. Chapter markers are well-structured, aiding navigation.

145 words

Title / Content Match

The title accurately reflects the content: the host recounts his experiment of granting an AI full access to his data, and the video explores the outcomes and implications.

Quality & Reliability

6/10

Discussion with practitioners and firsthand experience, but evidence is anecdotal and promotional bias is present. The video does not cite formal studies or independent data.

Chapters

Cited Sources

  • Qapten assistant — Product discussed as an example of an AI agent with user-controlled capabilities.
  • Michel Levy-Provençal LinkedIn — Guest and founder of Qapten.
  • Pierre Gaubil LinkedIn — Guest, entrepreneur and former founder of ScoreAm.
  • Carlos Diaz LinkedIn — Host of Silicon Carne and interviewer.

External References

Contribution & Novelties

The video offers a rare, detailed testimonial of a proactive AI agent operating autonomously on personal data, and contrasts it with a more conservative, user-controlled approach. It also highlights the emerging ’economy of laziness’ and raises ethical questions about delegation and cognitive sovereignty. The novelty lies in juxtaposing a cutting-edge American product (Instinct) with a French alternative (Qapten), providing a comparative perspective.

Pour aller plus loin :

  • Autonomous agent — Background on software agents that act independently.
  • AI alignment — Discussion on ensuring AI acts in accordance with human values, relevant to autonomous decision-making.
  • Digital sovereignty — Concept of controlling one’s data and cognitive autonomy, central to the debate.

109 words

Radar Profile

The profile shows high information quantity and moderate technical level, but reliability and consistency are average. The video provides rich content but lacks rigorous sourcing, resulting in a balanced yet not highly definitive radar shape.

Reliability 6/10