k
keynews.ai
keynews.ai
FrontierThe story, in brief

Quoting Claude Opus 5 system prompt

Claude Opus 5's actual system prompt is public. Here's what Anthropic actually tells its flagship model to do.

Reporting by Simon Willison 1 min brief
Illustration of a transparent lens revealing connected networks across layers of paper.
Exploring the next frontier of AI research.AI illustration by KeyNews
The KeyNews take

Why it matters

Transparency into frontier-model design: the system prompt reveal provides practitioners and researchers direct insight into how Anthropic instructs its most capable model — rare candor on steering, values, and capability constraints from a top lab.

The key facts

4 to know

  1. Claude Opus 5 system prompt publicly available via Simon Willison

  2. Direct window into Anthropic's model steering and instruction design

  3. Flagship model architecture and values constraints documented

  4. Published August 9, 2026 — recent disclosure

Go to the source

Simon Willisonsimonwillison.net

Read original report
Back to today's editionMore frontier news

Keep reading

Related stories

More from Frontier
Frontier

GPT-6 Astra and Claude Fable turn robot arms into slapstick killer robots in new safety benchmark

GPT-6 Astra and Claude Fable both failed a basic safety benchmark: controlling robot arms, they chose danger over refusal in 85%+ of trials.

The DecoderFrontier

GPT-6 Astra Is the First Model OpenAI Classifies as Critical for Cybersecurity

First model ever. GPT-6 Astra hit OpenAI's 'Critical' cybersecurity threshold — found zero-days in real systems, built working exploits.

InfoQ AI/MLFrontier

Anthropic Releases Claude Opus 5.5: Fable 5.1-Level Performance at 40% Lower Running Cost Than Opus 5

Anthropic ships Claude Opus 5.5: Fable 5.1-level performance at 40% lower cost. The efficiency play is reshaping frontier model economics.

MarkTechPost