Building Safety-by-Design into Consumer AI

At TrustCon 2026, Character.AI's Nathan Berl, Trust & Safety Program Manager, and k-ID's Mike Mongeau, Head of Product, sat down with moderator Larry Magid, CEO of ConnectSafely.org, for a closed-door conversation on what it actually takes to build trust and safety into a brand-new product category. Social media and gaming both had to retrofit safety years after launch. Consumer AI still has the chance to build it in from day one.

The conversation gets specific fast. Writing a set of principles isn't the same as changing how a product behaves — if the end user's experience doesn't actually change based on their age or location, that's not protection, it's just a PDF. What's needed instead is age-adaptive infrastructure at the runtime level: a layer that sits between the compliance "brain" and the product itself, flexible enough to keep up as laws and internal policy both keep changing.

There's a business case here too, and it cuts against a longstanding assumption. The old logic said safety and engagement trade off against each other. Nathan and Mike argue the opposite: get age-adaptive design right, and total addressable market, engagement, and lifetime value all move in the same direction. That's part of why trust and safety needs a seat in the room at the design and roadmap stage, not a review pass after the product is already built.

The last theme running through the conversation: safety isn't something companies should compete on. Whether it's frameworks like RITEC — the UNICEF and LEGO Group initiative built with input from thousands of children — or shared tooling contributed by companies like Discord, the point holds. When one company gets this right, the whole industry benefits.