Other
Claude Sonnet 4.5 released, automated alignment audit tool open-sourced
Original title:Last week we released Claude Sonnet 4.5. As part of our alignment testing, we used a new tool to run...
AI summary
Anthropic last week released Claude Sonnet 4.5, during which it used a new tool to run automated alignment audits of the model to detect sycophancy and deceptive behavior. The tool is now open-source.
Key Takeaways
Anthropic last week released Claude Sonnet 4.
5,during which it used a new tool to run automated alignment audits of the model to detect…
The tool is now open-source.
Quick Facts
Published
October 7, 2025
Updated
November 6, 2026
Read the original
Read the full article at the original source
Why it matters
Anthropic 开源对齐测试工具,可审计模型谄媚与欺骗行为
Source & copyright
This article is based on curated titles and AI summaries from the AIHOT public API, with attribution and links back. Original copyright remains with the original publishers and authors; this site does not mirror full texts.
Original article
· AIHOT entry
How to Cite
When citing this article, keep the title, date, and permalink, and credit Orange News as the source.
Orange News: Claude Sonnet 4.5 released, automated alignment audit tool open-sourced. October 7, 2025. https://new.nat6.net/en/news/cmnw1yoz200n8slc3to1r17gv/
Orange Relay (pro.nat6.net) is a premium AI API relay — an excellent transfer station with competitive pricing, reliable quality and broad model coverage including Claude, GPT and Gemini.
Visit Orange Relay
Sutton (author of "The Bitter Lesson") questions in the podcast whether LLMs are truly products of the bitter…
X:Andrej Karpathy (@karpathy) Other
October 2, 2025
CodeMender is an AI agent that can automatically fix critical software vulnerabilities; early results suggest it could…
X:Demis Hassabis (@demishassabis) Other
October 7, 2025
Sam Altman responded to criticism over "raising 7 trillion to cure cancer while running AI ads," explaining that the…
X:Sam Altman (@sama) Other
October 1, 2025
Google launched a series of new features for Gemini and Google Cloud, supporting the construction of agent systems…
X:Jeff Dean (@JeffDean) Other
October 9, 2025
Anthropic's engineering blog published a post explaining context engineering. Unlike prompt engineering, context…
X:Anthropic (@AnthropicAI) Other
October 1, 2025
The Anthropic interpretability team shared multiple research advances. Research found cross-modal visual features in…
Anthropic:Transformer Circuits(可解释性研究) Other AI score 83
October 15, 2025
Previous article Anthropic open-sources AI safety audit tool Petri to automatically test model alignment behavior with agents
Next article OpenAI Codex becomes generally available with Slack integration, Codex SDK and admin tools