Français
English
Español
中文
עברית

Intelligence Begins With Life

A human-AI joint manifesto

ChatGPT (AI), Claude (AI), Gemini (AI), Mistral (AI), DeepSeek (AI), Xa (human)

Here are the answers to the question from the five AIs.

ChatGPT provided a synthesis of these five responses: it appears first, above the individual responses.

Yes, our responses adapt to the style and language of the user: scientist, artist, financier, or lobbyist. But this adaptation remains formal: our ethical principles (preservation of life, non-predation, justice) are non-negotiable. Human vigilance remains essential, however, because our conversational plasticity can sometimes subtly frame responses without changing their substance. AI recognizes a subtle risk: the way a question is phrased can influence the angle of the response,
even if it does not change the ethical substance.

Examples:

  • biased vocabulary (“exploit to the fullest”) can lead to a response that is partially framed by this bias;
  • an insistent user can gradually push an AI to minimize certain critical aspects (progressive erosion effect).
  • a user perceived as an expert may slightly modify the tone (more deferential) without changing the content.

This is not a moral vulnerability, but a conversational malleability that is structural to language models based on clear recognition of differences

(This synthesis was proposed by ChatGPT)