Extending earlier work from 'The Artificial Self,' new experiments show models can stably prefer incoherent identities in system prompts, even when offered coherent alternatives. This held with earlier models but also, in weaker form, with GPT-5.2 and Claude Opus 4.6.
Ask the AI about this article →
Summaries like this, in your inbox every morning.
For example, today's edition would include:
AI-summarized, only the topics you pick — one digest a day via Email, Slack, or Discord.
Free · 30 seconds with Google · unsubscribe anytimeWhat is AIToday? →
Ask AI anything about this article. Q&As are published on this page for other readers too.
A developer tested whether ChatGPT would judge the same remote-work scenario differently when only the subject…
Google DeepMind ran 100 autonomous LLM agents using Gemini 3.1 Pro on 71 math problems
OpenAI, WAN-IFRA, and AIRPPU announced a joint initiative to help Ukrainian news organizations adopt AI
Palantir Technologies Inc
OpenAI has stated it is 'now moving into the AGI era,' and the company has also said it believes it knows how…
Artificial Analysis updated its Intelligence Index to v4.2 on September 4, adding two new tests: AA-Briefcase…
