Project summary

Every AI safety tool on the market detect harmful content and messages against rigid rules. But emotional harm does not show up in one message. It builds quietly over time - user-relative, relational and cumulative, thus cannot be detected by static, rule-based methods.

I replace static methods with dynamic, pattern-based methods - converting the back-and-forth human-AI interactions into sound and music, a tangible signal to be measured over time, so emotional harm can be detectable before it escalates.

What are this project's goals?

Global regulations are already demanding emotional AI safety. For instance, California SB 243 banned engagement-maximising design and required third-party audits, and three other jurisdictions (US State, EU and China) are all pushing similar rules.

So my goal at this stage is:
To build an engine that translates the human-AI interaction's emotional dynamics into sound and music signals, so affective harm patterns like over-attachment, manipulation or isolation become measurable and detectable.

How will this funding be used?

At 10,000 min, it gets me a working MVP and lets me start testing it against 100+ real interactions in conversational AI (i.e., companion AI) to see what needs to be modified.

At 55,000 max, it funds me full-time for six months to turn this into affective guardrails for AI enterprises.

Who is on your team? What's your track record on similar projects?

AI Governance: I worked on AI governance at Amazon, focusing on EU AI Act, Directive on Platform Work and GDPR. So, I have sit inside the audit process this project needs to eventually pass.

Behavioral Science: I was the behavioral science researcher at King's College London. I researched how ASMR affect people with higher neuroticism. So, I have already studied how audio-based signals interact with emotional vulnerability.

How much money have you raised in the last 12 months, and from where?

Zero