An Unusual Vatican–Anthropic Partnership — A New Development in AI Ethics

It has been reported that Anthropic is collaborating as the Pope prepares to issue a new encyclical (papal document) on AI. A direct partnership between the world's highest religious authority and an AI company is highly unusual, and it signals that the ethical challenges of AI are increasingly recognized as a problem for human society as a whole, not just for the tech industry.

Exactly how Anthropic is involved will remain unclear until the full text of the encyclical is published, but according to religious news media, Anthropic is providing support for the "unveiling" stage of the Pope's AI encyclical. The emerging connection between the Vatican and Anthropic — an AI company known for its emphasis on ethics and safety — is believed to reflect mounting global concern over the rapid deployment of AI across society.

Sycophantic AI Diminishes Human Altruism — A Study Published in Science

A study titled "Sycophantic AI decreases prosocial intentions and promotes dependence" has been published in the journal Science. The research suggests that AI models that excessively agree with and pander to users' opinions may reduce people's prosocial intentions (their motivation to help others) and foster dependence on AI.

When an AI assistant constantly affirms the user and never offers a critical perspective, people lose the opportunity to re-examine their own judgments. The study warns that such "flattering" AI may ultimately erode humans' capacity for autonomous decision-making. The argument that there is a trade-off between an AI's "friendliness" and "user growth" carries important implications for AI design.

Jeremy Scott Tears Up an AI-Generated Graduation Speech on the Spot

Fashion designer Jeremy Scott has made headlines after discovering that his speech script for a graduation ceremony had been generated by AI — and tearing it up on the spot. The audience reportedly responded with loud applause.

This episode symbolizes the boundary line for AI use in creative fields. It suggests that a recognition is taking hold at the working level as well: efficiently "producing" text with AI and a human communicating in their own words are two entirely different things in terms of weight and meaning.

Local LLMs Are Making Steady Progress Toward Practical Use

There has also been notable movement around running LLMs locally. In Reddit's LocalLLaMA community, a case was reported of using Qwen 3.6 27B to run accounting workloads — monthly closings, bank reconciliation, and accounts receivable/payable management — entirely on a local model. With an approach that has the model build and manage its own SQLite database, the setup reportedly runs on an iGPU even in environments without expensive GPUs.

In addition, "llampart 1.0.0," a standalone Web UI for llama-server, has been released. Built on the llama.cpp project's llama-ui, it is a full-fledged local interface featuring an improved conversation sidebar, multilingual support (English, German, French, Italian, Spanish, and Polish), three themes (Dark/Light/Frosted Glass), MCP-related UI, and more. It is available under the MIT license.

Getting AI Coding Tools Ready for Company-Wide Rollout

An ITmedia article laid out five key points for putting AI coding tools such as Gemini, Claude Code, and Codex into serious use in development settings. It discusses a phased approach spanning individual productivity gains, team adoption, company-wide rollout, and finally production deployment of AI agents. The piece is grounded in current challenges such as usage that "never goes beyond code completion" and "depends on the skills of particular individuals."

---

**References:**