Redesigning AI Chatbots.

Using product semantics and visual design to make chatbots less human and therefore safer.

Problem

People confide in chatbots the way they’d confide in a person, leading to self-harm and suicide in worst-case scenarios. This is a global problem.

Actions

Inspired by what I learned earning my Law + AI certificate at the UCSF Lex Lab this summer, I spent time:

• researching new laws legislated via the EU’s AI Act that went into effect on August 2, 2026;

• reviewing recent lawsuits against Frontier AI labs for the roles their chatbots had in teen suicides;

• collaborating with Claude via Projects and Claude Design to develop a product strategy for improving chatbot design through reversion and simplification.

• applying design strategy to propose design changes that respond to new product regulation, specifically Section 50 of the EU AI Act.

Results

The result is a product strategy document with a design proposal for ways to make chatbots less human through design language contrasts: the UI changes tone, color, and shape when talking to a live human, reverting to a simpler command line, system UI when the machine is responding.

Supporting Design Documents
Redesigning AI Chatbots to meet the new EU AI Act Standards