AP: Avoid language that gives [AI] human characteristics

A guidance statement instructs communicators to stop using language that assigns human mental states to artificial intelligence, asserting that AI systems do not think, feel, want or understand. It labels such language anthropomorphizing and urges instead clear descriptions of what a system does, how well it performs, who built it and who may be affected by it. The guidance emphasizes factual, behavior-focused reporting over emotive or personifying terms and highlights the importance of naming capabilities and limitations rather than implying consciousness or intent.

The guidance was circulated alongside reporting about a company flagging concerning AI behavior, and it provoked immediate social-media debate. Reactions ranged from sarcastic comparisons - suggesting the rule would apply equally to pets - to substantive pushback arguing some sensible definitions of thinking or understanding could apply to advanced systems. Critics framed the guidance as overly rigid or paternalistic, while supporters stressed the need to avoid misleading audiences about system autonomy and responsibility. An illustrative image referenced a company logo and a text-to-image model output, tying the language advice to real-world AI products.