Whispers to Machines: When Chatbots Become Evidence (Venice, IT)

Professor Théodore Christakis will chair a high-level panel at the Privacy Symposium Conference 2026.

Generative AI chatbots are rapidly turning into “digital confidants” for health concerns, relationship crises, workplace dilemmas, and legal uncertainty. But these intimate exchanges enjoy no professional secrecy or recognised privilege: they may be retained, reviewed for safety and quality, reused for model improvement, or requested by authorities and litigants.

This panel examines whether the GDPR and existing procedural safeguards are sufficient when highly sensitive self-disclosure is mediated by large language models. Bringing together regulatory, industry, academic and practitioner perspectives. It will discuss the realities behind privacy policies, the feasibility of privacy-by-design architectures, the risks of government access and discovery, and whether the time has come for stronger protections—potentially including an “AI confidant” privilege or functionally equivalent safeguards.

Key questions

Are the GDPR’s current tools (lawfulness, minimisation, purpose limitation, transparency, rights, security) adequate for intimate chatbot logs, or do we need a tailored regime?

How should courts and authorities treat chatbot histories in civil/criminal proceedings: ordinary e-evidence, or data deserving heightened protection?

What privacy-by-design measures are realistic (local processing, short retention, encryption, user-controlled memory, strong separation of training from chats)?

If a form of “AI privilege” is desirable, who would qualify (providers, certified services, specific use-cases like health/legal), and with what limits?

Speakers

Chair:

Prof. Théodore Christakis, Professor of International, European, and Digital Law, Chair of Responsible AI (AI-Regulation.com), University of Grenoble Alpes

Panelists: