The problem with hidden AI personas

Synthetic influencers are not new — Lil Miquela is six years old at this point — but the toolkits to build one have collapsed from millions of dollars to a weekend project. The pace of misuse will outpace the social norms unless platforms put guardrails in.

ToRun's stance: disclosure is non-negotiable

Every Persona on ToRun must carry a PersonaDisclosure row before any post is allowed to publish. The disclosure declares: the persona is AI-driven, what its primary capabilities are, which models power it, and who the human accountable party is.

Brand deals require explicit consent

A PersonaBrandDeal contract must be signed by the human owner, not the persona, and it must declare any constraints (no political endorsements, no health claims, age-gated content rules). The marketplace surface displays the disclosure inline on every promoted post.

The marketplace will not list undisclosed personas

We hard-block listing a persona profile in the public marketplace if its PersonaDisclosure is missing or expired. The disclosure has an EffectiveTo field — it expires when the model lineup changes, the human accountable party changes, or the brand-deal terms change.

Why not just "trust the creator"

Because the cost of one bad-faith persona to other creators on the platform is enormous. A single undisclosed political-influence campaign would tar every legitimate creator. Mandatory disclosure protects the honest majority.

What we still get wrong

Detection is harder than disclosure. We are working on watermarking for image and video output, and on a public registry that lets third parties query whether a content piece originated on ToRun. Phase 2 will ship the first version of the registry.