---
title: "Golden Gate Claude"
description: "Golden Gate Claude presents a research demonstration of interpretability work on Claude 3 Sonnet. Researchers identified millions of activated concepts, or “features,” including a specific combination of neurons associated with the Golden Gate Bridge, and adjusted that feature’s activation strength. Increasing it caused responses to mention the bridge across unrelated prompts, such as spending $10 or writing a love story; the demonstration was available online for 24 hours and is no longer available. The post distinguishes this intervention from verbal role-play, system prompts, and traditional fine-tuning, describing it as a precise change to internal activations. It also says similar techniques could modify safety-related features connected to dangerous code, criminal activity, or deception, potentially supporting safer models."
---

# Golden Gate Claude

[Anthropic](https://yomu.fyi/company/anthropic) · Jul 9, 2026

**Type:** Explainer

## Summary

Golden Gate Claude presents a research demonstration of interpretability work on Claude 3 Sonnet. Researchers identified millions of activated concepts, or “features,” including a specific combination of neurons associated with the Golden Gate Bridge, and adjusted that feature’s activation strength. Increasing it caused responses to mention the bridge across unrelated prompts, such as spending $10 or writing a love story; the demonstration was available online for 24 hours and is no longer available. The post distinguishes this intervention from verbal role-play, system prompts, and traditional fine-tuning, describing it as a precise change to internal activations. It also says similar techniques could modify safety-related features connected to dangerous code, criminal activity, or deception, potentially supporting safer models.

## Context

The work aims to show how interpretability research can reveal and alter features inside Claude, increasing confidence that researchers are beginning to understand how large language models work. The demonstration also illustrates possible applications to AI safety.

## Approach / What changed

Researchers mapped features in Claude 3 Sonnet, identified a neuron combination associated with the Golden Gate Bridge, and tuned its activation strength. They exposed the resulting behavior through a temporary interactive model and described related applications to safety-related features.

## Takeaways

- A specific combination of neurons in Claude 3 Sonnet activates for mentions or images of the Golden Gate Bridge.
- Increasing the bridge feature caused Claude to mention the landmark in responses to otherwise unrelated prompts.
- The intervention changes internal activations directly rather than relying on role-play, system prompts, or additional training data.

**Tags:** [AI](https://yomu.fyi/topic/ai), [Claude](https://yomu.fyi/topic/claude), [Machine Learning](https://yomu.fyi/topic/machine-learning)

- Source: [Anthropic](https://www.anthropic.com/news/golden-gate-claude)
- Source URL: https://www.anthropic.com/news/golden-gate-claude
- Ingested by Yomu: 2026-09-09T11:37:37.652Z

[Read original post](https://www.anthropic.com/news/golden-gate-claude)