Anthropic releases two policy proposals on how governments should address catastrophic risks and manage labor market disruption from advanced AI systems
and the gap between the two is becoming the central challenge of the technology. In his latest essay, our CEO Dario Amodei lays out how to close it. We're launching three new initiatives to support the
Context & Ripple Effects
Anthropic has been building a policy case for government action since its 2024 regulatory principles, and has since paired that agenda with Dario Amodei's warnings about potentially severe entry-level white-collar job disruption. Its recent policy work has also tied frontier-model safeguards to US strategic competition and export controls.
This release brings catastrophic-risk governance and labor-market adjustment into one policy frame. Anthropic is no longer arguing only for controls on powerful models; it is also asking governments to prepare for the distributional consequences of deployment.
First-order effects
- Governments and policymakers now have a more explicit Anthropic-authored agenda spanning frontier-model risk management and worker disruption, reinforced by three supporting initiatives.
- Anthropic further positions itself as a participant in setting expectations for advanced-model oversight, alongside its call for third-party testing and transparency for cyber, bio, and autonomy risks.
Second-order effects
- Other frontier-model developers may face greater pressure to state comparable positions on independent evaluation, transparency, and the employment effects of their systems rather than treating safety and labor policy separately.
- The proposals sharpen the political trade-off around AI governance: safeguards and workforce preparation can become part of the case for deployment, while critics may continue to characterize company-backed rules as favoring incumbents.
Third-order effects
- If policymakers adopt a combined approach, frontier AI governance could evolve from model-safety rules alone toward a broader regime linking access to powerful systems with labor-market transition planning.
- The durability of that approach will depend on whether requirements can be applied across developers without locking in the largest firms—a tension already visible in criticism of Anthropic's policy advocacy.
The trend: Frontier AI policy is shifting toward treating catastrophic capability risks and employment disruption as interconnected governance problems rather than separate debates.