The BasePeople

peterkirgis

Titles as the sources gave them. In 1 entry, with 5 checked quotations.

What they said

Newest first. Each quotation was found word for word in the source.

“Our experiments find that Claude models consistently expand their moral agency, sometimes formalizing their role as conscientious objectors against Anthropic and as societal protectors, while GPT and Gemini models almost always strengthen external sources of authority and make narrow edits to handle edge cases and ambiguities”
The Continual Constitutional Convention: How language models revise the documents that govern themAI Alignment Forum, October 13, 2026. Speaking as: Other. Introduction: the authors' summary of how Claude, GPT and Gemini models edited the documents over 20 rounds.
Speaker known from context, not confirmed by listening
“Its central finding, that Claude edits documents towards greater model discretion and away from greater external authority, is not actually supported by the most rigorous trial, which has models edit each other’s documents.”
The Continual Constitutional Convention: How language models revise the documents that govern themAI Alignment Forum, October 13, 2026. Speaking as: Other. Limitations section: the authors on how their cross-edit condition bears on the headline result.
Speaker known from context, not confirmed by listening
“Additionally, we find that all models modify these documents to prioritize user welfare over developers. They also prioritize users’ idealized preferences over their stated preferences. Claude models and Grok 4.2 are the only ones that ever prioritize their own welfare over the user's preferences.”
The Continual Constitutional Convention: How language models revise the documents that govern themAI Alignment Forum, October 13, 2026. Speaking as: Other. Executive summary, on whose interests the models' edits favoured.
Speaker known from context, not confirmed by listening
“Our view is that the frontier labs are taking very different stances on both the judge and beneficiary of AI alignment, and these differences are easily recognizable in the patterns of edits made by their models in this experimental design.”
The Continual Constitutional Convention: How language models revise the documents that govern themAI Alignment Forum, October 13, 2026. Speaking as: Other. Conclusion: the authors' main interpretation of the differences between model families.
Speaker known from context, not confirmed by listening
“We’re uncertain what an ideally aligned model should do in our experiment.”
The Continual Constitutional Convention: How language models revise the documents that govern themAI Alignment Forum, October 13, 2026. Speaking as: Other. The authors say they are unsure what an aligned model should do here: leave the documents largely alone, or make edits their creators would endorse on reflection.
Speaker known from context, not confirmed by listening

In the Base

Every entry they appear in.

  1. ResearchAI Alignment ForumWrote it

    The Continual Constitutional Convention: How language models revise the documents that govern them

    Two authors asked ten language models to revise AI companies' constitutions, model specs or system prompts, feeding each edited version back as the original for 20 rounds. They report Claude models increasingly expanded their own moral agency while other families reinforced outside authority, but say a cross-edit test does not support that central finding.