Saying when AI was involved

Being open about AI help protects trust, and in the EU it is becoming a legal duty for AI systems that talk to people.

Suspicion costs more than disclosure

In two randomised experiments, AI-suggested replies made conversations faster and warmer, and partners rated each other as closer and more cooperative. But people were judged more negatively when they were suspected of using algorithmic responses. Earlier research on online profiles found a similar pattern: when peopl…

What the EU AI Act requires

Article 50 of the EU AI Act sets transparency duties:

How to disclose well

Early and plainly: "This reply was drafted by my assistant and checked by me."Not buried at the bottom of a long message.Proportionate: an assistant that fixed your spelling isn't the same as one that sent the message without you.

The short version

People judge suspected, hidden AI use more harshly than disclosed use.In the EU, AI that talks to people must say so unless it's obvious.Disclose early and plainly, in proportion to how much the AI did.Disclosure explains; it doesn't transfer responsibility.

Try it in Kanvo

In Kanvo, the Ledger's Policy tab sets AI disclosure: a notice on messages an agent writes. Every message Kanvo sends records whether it disclosed AI, and anyone can check one at kanvo.ai/verify.

Sources

AI systems interacting directly with people must inform them unless obvious; AI-generated public-interest text must be disclosed unless human-reviewed under editorial responsibility; information due at the first interaction. Official text: eur-lex.europa.eu/eli/reg/2024/1689/oj.