Anthropic releases two policy proposals on how governments should address catastrophic risks from powerful AI models and how to prepare workers for AI's impact
AI is advancing at exponential speed, and the policymaking process was built for a slower world. — We are sharing two policy proposals to prepare for AI progress.
Context & Ripple Effects
Anthropic has moved from a broad call for near-term AI policy principles in 2024 to more specific positions on export controls, model-security risks, and potential slowdowns or pauses for highly capable systems. The new proposals extend that policy campaign into two linked areas: catastrophic-risk governance and labor-market preparation.
The intervention arrives as OpenAI has also advanced proposals for the economic consequences of superintelligence, making AI-lab policy advocacy itself a more visible part of competition over how advanced AI will be governed and how its gains are distributed.
First-order effects
- Governments receive a more concrete Anthropic framework for addressing high-consequence model risks and preparing workers for disruption from advanced AI.
- Anthropic strengthens its position as both an AI supplier and an active advocate for rules governing deployment, safety thresholds, and transition planning.
Second-order effects
- Other frontier-model developers face greater pressure to articulate comparable positions on safety governance and labor adjustment, rather than treating those issues as separate from product development.
- Policy debates may increasingly tie access to or deployment of powerful models to obligations beyond technical safety, including worker-transition measures and public economic protections.
Third-order effects
- If major labs continue publishing detailed policy agendas, frontier-AI governance could be shaped jointly by governments and the companies building the systems, raising enduring questions about accountability and whose interests set the rules.
- Catastrophic-risk controls and labor-market policy may become a single governance agenda for advanced AI: managing both the hazards of capability and the distributional effects of adoption.
The trend: Frontier AI companies are increasingly seeking to define the safety and economic-policy framework for the systems they are racing to build.