Resources: Foundations of AI Governance

Exercises: Foundations of AI Governance

Resources: Proposals to Govern AI

Exercises: Proposals to Govern AI

Next steps: Programs

Resources: The AI Policy Toolkit

Exercises: The AI Policy Toolkit

Resources: How AI is Governed Today

Exercises: How AI is Governed Today

Resources: Governance at Frontier AI Companies

Exercises: Governance at Frontier AI Companies

In this unit, we'll zoom in on how frontier AI companies are approaching safety and responsible development. These companies, at the cutting edge of AI capabilities, are experimenting with novel approaches to corporate governance and self-regulation.
A key focus will be Responsible Scaling Policies (RSPs) - frameworks that labs use to evaluate and manage risks as they advance AI capabilities. These policies represent an attempt at industrial self-regulation, but their effectiveness and limitations are hotly debated. We'll examine both traditional and nontraditional corporate structures, exploring how different organizational models might influence safety outcomes.
Remember that corporate governance frameworks reflect specific organizational priorities and constraints. As you analyze these approaches, consider the inherent tensions between innovation, safety, and commercial pressures, and how different organizational structures might help or hinder managing these tradeoffs.

By the end of the unit, you should be able to:
\- Analyze how frontier AI labs govern themselves through both organizational structure and internal policies
\- Evaluate the strengths and limitations of Responsible Scaling Policies as a safety tool
\- Explore other ways frontier AI labs play a role in AI safety


_Spend 20-30 minutes on this exercise_
Review the responsible scaling policy of one of the following frontier labs:
- [OpenAI: Preparedness Framework](https://cdn.openai.com/openai-preparedness-framework-beta.pdf)
- [Anthropic: Responsible Scaling Policy](https://assets.anthropic.com/m/24a47b00f10301cd/original/Anthropic-Responsible-Scaling-Policy-2024-10-15.pdf)
- [DeepMind: Frontier Safety Framework](https://storage.googleapis.com/deepmind-media/DeepMind.com/Blog/introducing-the-frontier-safety-framework/fsf-technical-report.pdf)
- [Meta AI Safety Policies](https://transparency.meta.com/en-gb/policies/ai-safety-policies-for-safety-summit)
    - Meta has not released a formal RSP. Instead, they submitted a list of safety policies to the UK Bletchley Summit.
Then, answer the following questions:
1. What does this policy do well?
2. Where could this policy improve?
3. Will this policy effectively mitigate major risks? If yes, why? If not, what makes it insufficient?


AI Governance Fast-Track

Exercises: Governance at Frontier AI Companies

Exercises