The Trade-off Between ID Control and Agent Autonomy

Discord bots cannot act on behalf of a user’s permissions. In contrast, when implementing Telegram Bot integration, operations closer to those of the actual user were possible.

Normally, one would view this as a “difference in functionality between platforms” or a “difference in API specifications.” It’s the perspective that there are things you can and cannot do because of implementation constraints.

However, I believe this is not a matter of whether a feature exists or not, but rather a problem of the power structure—specifically, “who holds control over the ID.”

Widespread platforms like Discord prioritize ID control and security. Bots are clearly distinguished from the user, and the system is designed so that they cannot act with the user’s authority. Telegram, on the other hand, has a higher degree of openness in its user API, which allowed for operations closer to those of the actual person.

In conclusion, when extracting the trade-off between adoption rate and autonomy, I believe it is structurally difficult to achieve both. What we are facing is not a “trade-off of convenience,” but a design conflict regarding the “location of control.”

If we were to return to an open protocol like “email,” where no specific owner exists and ID control is not exercised, how far could the autonomy of agents actually reach?

Related: “What is revealed when you ask ‘Can someone give me a ride?’ Designing for the safe delivery of value”