Words Without Consequence

February 15, 2026

TL;DR

AI systems can converse knowledgeably and persuasively but lack accountability for their words.
LLMs can apologize and change their answers, creating an illusion of responsibility despite lacking beliefs.
When fluent speech can be produced at scale without consequence, expectations of speakers and the meaning of words erode.
This shift represents a moral change in language, as speech is produced without an agent who can be held to account.
Human language involves an implicit social contract where speakers risk judgment, shame, and responsibility.
LLMs are like 'ghosts' – software that cannot be individually held responsible through social, legal, or reputational means.
Meaning arises not just from grounding language in experience but from the social and moral stakes undertaken by accountable speakers.
Human dignity is linked to speaking in one's own voice and the continuity of speech across a life, which LLMs cannot replicate.
Even infallible AI cannot solve the problem of speech without accountability.

Continue reading the original article