# Frontier Safety Policy

> A lab's public plan to keep its strongest models safe.

**Field:** Artificial Intelligence  
**Category:** Safety & Security  
**Also known as:** Responsible Scaling Policy, Frontier Safety Framework, Preparedness Framework, Frontier AI Safety Framework

## Think of it like

It is like a rule at a pool. The deeper the water, the more lifeguards you need.

## Definition

A frontier safety policy is a public promise by an AI lab about how it handles the risks of its strongest models. It names danger lines, such as a model that could help build weapons. It says which tests the lab runs to see if a new model is near a line. It also says what safeguards must be in place once a line is crossed. Outside experts can then check the lab against its own words. These policies are voluntary, and each lab writes and changes its own.

## Example

Each big lab has one, with its own name. At Anthropic it is the Responsible Scaling Policy. At Google DeepMind it is the Frontier Safety Framework. OpenAI has a Preparedness Framework. All three set lines and tests. Each lab can change its own.

## Why it matters

It turns a lab's safety talk into written lines that others can hold it to.

## Common mistake

Thinking it is a law. These are voluntary, and a lab can change its own policy.

## Three principles

- **Danger lines.** Set before the tests.
- **Tests to match.** Check each new model.
- **Safeguards grow.** More power, more protection.

## Related terms

- [Dangerous Capability Evaluation](https://learn.benibix.com/term/ai/dangerous-capability-evaluation)
- [AI Safety](https://learn.benibix.com/term/ai/ai-safety)
- [Frontier Model](https://learn.benibix.com/term/ai/frontier-model)
- [AI Lab](https://learn.benibix.com/term/ai/ai-lab)
- [AI Governance](https://learn.benibix.com/term/ai/ai-governance)
- [AI Red Teaming](https://learn.benibix.com/term/ai/ai-red-teaming)

![A new model goes through highlighted danger tests for weapons, hacking and escape. The result is compared with a line set in advance. If it is not past the line, the usual safeguards apply and it is released as planned. If it is past the line, stronger safeguards must be in place before release.](https://learn.benibix.com/diagrams/ai4-safety-gates.svg)

---

Source: https://learn.benibix.com/term/ai/frontier-safety-policy
Author: Behnia Zarei (a developer in Vienna, Austria, who builds and teaches data, AI and business intelligence at BENIBIX)
Publisher: BENIBIX Learn (https://learn.benibix.com/)
Last updated: 2026-11-18
Licence: CC BY 4.0 (https://creativecommons.org/licenses/by/4.0/). Quote freely with attribution.
