Issue #29, September 25, 2026
Muse, the democratisation of personal assistants and the privacy challenge
Sébastien Louradour
9/25/20262 min read
Muse, the democratisation of personal assistants and the privacy challenge
Meta has launched the app anyone would have expected when GenAI started to emerge a few years ago. A personal assistant, call it a personal butler, or an executive assistant, able to complete everyday tasks, such as booking a flight or a restaurant, or purchasing something for you. The dedicated app has become an instant success: it's the #1 download on the US App Store and Meta's stock has surged too. This radical approach to building AI for everyone ultimately raises new questions, including what level of personal data, including sensitive data, will be required for the AI assistant to operate meaningfully. In the EU, the GDPR framework likely explains why Muse hasn't yet landed in the old continent. The data privacy issue is about to play out very loudly between tech giants too. At WWDC in June, Apple insisted privacy was at the core of everything Siri would do, thanks to on-device processing and its Private Cloud Compute servers. To work properly, AI assistants such as Muse (we can expect Siri could ultimately become a product with the same features) need a very deep knowledge of ourselves: (i) formal information, such as credit cards, passports, addresses, etc. and (ii) informal information: our holiday habits, our personal preferences for style, way of life, social and economic categories, etc.
So comes the question of how regulation can adapt to such product evolution, and how the product roadmap will be impacted by ongoing regulations. European regulation, starting with GDPR, will heavily constrain the capacity for Muse to operate freely. Right now, GDPR is built around purpose limitation and data minimisation, with explicit consent as the default gateway for sensitive data. But an AI assistant has a purpose evolving constantly based on the user's needs, and sensitive data is ubiquitous in our daily lives: religion, sexual orientation and health are embedded in purchase and activity choices, making it very complex to provide a service without having to process such data. So the question is how and whether an AI assistant can process sensitive data and what practical framework can enable it, while maintaining a level of control and a seamless user experience. The ongoing discussion of the Digital Omnibus has mostly focused on the use of sensitive data for model training, and the text so far only carves out a narrow exception for sensitive data unintentionally present in training sets, rather than building a framework that would enable AI assistants in the EU. The Commission’s Digital Fitness Check could be the right vehicle to address the gap left by the Omnibus: the lawful processing of sensitive data by AI assistants in deployment, not just in training.
Muse represents an opportunity to democratise access to services once only provided to the wealthiest. Now the question is how this vision can materialise in Europe. The power of user experience, and the capacity to ship a vision that has propelled Big Tech through the roof, could play out again. But even with popular success, privacy will represent the next battleground, both between two competing visions (Meta vs Apple) and between the EU and AI labs.