strategy tag
RSP-style commitments.
Responsible scaling policies; labs commit to capability-tied safety
stated endorsers
8
no opposers yet
profiled endorsers
1
248 on the board total
endorser mean p(doom)
18%
n=1 · median 18%
quotes by endorsers
8
just for this tag
People on the record.
8Anthropic Policy Team (RSP authors)
Anthropic responsible scaling policy author
Helped draft Anthropic's Responsible Scaling Policy, a tiered capability-and-safety commitment regime.
The Responsible Scaling Policy ties safety commitments to specific capability thresholds. If evaluations show we have crossed a threshold, we pause deployment until mitigations are in place.
Dario Amodei
CEO of Anthropic; 'Machines of Loving Grace' author
Championed Responsible Scaling Policies: capability thresholds trigger progressively stronger safety commitments.
If we put enough effort into solving these problems, a truly amazing, hopeful future could be available.
Context: Opening framing of Machines of Loving Grace, which argues powerful AI could compress 50–100 years of biological progress into 5–10.
Helen King
DeepMind VP of Research; responsibility lead
Leads DeepMind's Frontier Safety Framework, a capability-tied safety commitment regime.
The Frontier Safety Framework commits us to evaluations at defined capability thresholds, with mitigation protocols triggered accordingly.
Joscha Romeike
Anthropic policy team
Helps Anthropic engage with European regulatory bodies on Responsible Scaling Policy and the EU AI Act.
Capability-tied safety frameworks need to be legible to multiple jurisdictions. That's a translation problem we work on.
Lila Ibrahim
DeepMind COO; AI ethics governance
Argues responsible scaling, publicly committed safety measures triggered by capability thresholds, is how frontier labs must operate; helps shape DeepMind's framework.
Our Frontier Safety Framework defines protocols for responding to elevated risks tied to powerful capabilities. Safety must scale alongside capability.
Michael Page
Anthropic policy team
Helps operationalise Anthropic's Responsible Scaling Policy and represent it to governments.
Capability-tied safety commitments must be credible to regulators, not just internal talking points.
Sandhini Agarwal
OpenAI policy researcher
Argues policy and safety teams need to be embedded in deployment decisions, not consulted after the fact; this is the structural reason RSPs and preparedness frameworks have value.
We discuss the social and economic implications of deploying language models. The deployment timeline cannot be separated from the policy work that surrounds it.
Thomas Larsen
Center for AI Policy founder; AI 2027 co-author
Argues responsible-scaling-policy-style commitments need legal teeth: voluntary lab pledges are not adequate when commercial pressure conflicts with safety.
Voluntary RSPs are necessary but not sufficient. We need legal frameworks that make responsible scaling commitments enforceable when economic pressure pushes labs the other way.