Can slowing AI development resolve who stops deployed systems?
Pace measures like embedded evaluators and capability checkpoints can govern how fast capabilities advance, but do they address the separate problem of intervention authority after deployment? The question asks whether the same tools that slow development can also handle deployed-system governance.
The discussion frames the current debate this way: "The ensuing debate has focused principally on the pace of development. Dario Amodei's proposal would slow the frontier through embedded evaluators and coordinated capability checkpoints. Such measures govern the conditions under which capabilities advance. They do not resolve who may intervene when a deployed system causes harm, or how that intervention should proceed."
The distinction is about the object and the time. Pace measures act on development: whether, how fast and under what checks a capability gets built. The paper's question is about a system already out. Two questions survive even a perfectly implemented slowdown, who has the authority to intervene and by what procedure, because both arise after the capability exists and is deployed. The paper does not call pace measures wrong. It concedes in the next sentence that slower development may reduce risk (Does slowing AI development actually prevent system failures?).
What the excerpt relays and what it leaves out. The proposal is described in one sentence and cited to footnote 243, which the excerpt does not open, so nothing can be said here about it beyond that sentence, and its author may describe it differently. The excerpt also does not say whether a coordinated capability checkpoint could double as a trigger for interrupting a deployed system; it treats the measures as acting on advance only. Whether that is fair to the proposal is not something these paragraphs settle.
Where the vault already stands. Can regulation keep pace with AI's rapid evolution? asks how rules can keep pace with a moving target. This paper's reply to the pacing frame is to ask a different question, what authority exists after release, and that reading of the contrast is the vault's.
Inquiring lines that read this note 9
This note is a source for these research framings, grouped by the broader line of inquiry each explores. Scan the bold lines of inquiry; follow any specific question forward.
What determines whether AI system errors remain visible and contestable?- How do you stop an AI system once it is already deployed?
- What counts as a successful stop or intervention on a deployed AI system?
- How often do deployed AI systems actually get stopped when they cause harm?
- What authority should exist to stop an AI system once deployed?
- How does coordination governance shift the hard problem from capability itself?
- Why do legal and institutional stops matter more than technical ones?
- How do intervention rules change when slowing pace does not prevent harm?
- Who actually has the authority to stop a deployed AI system?
Related concepts in this collection 4
This note in its neighbourhood — explore the map, then jump to a related concept in the list below.
Click a node to walk · click center to open · click Open in graph to see this note in the full knowledge graph
-
Does slowing AI development actually prevent system failures?
Explores whether pace constraints reduce risk enough to eliminate failure in tightly coupled AI systems. Matters because the debate often conflates risk reduction with failure prevention.
the paper's premise for why the remaining questions matter
-
How do we stop AI systems once they are already deployed?
Current AI governance focuses on what gets released, but deployed systems create a separate problem: who has the power to halt them and how? This gap may be where governance frameworks are now failing.
the thesis this distinction supports
-
Does agent capability matter more than coordination infrastructure?
As AI agents take on economic and social roles, what actually limits their effectiveness: the raw reasoning power of the model itself, or the systems that let them coordinate, stay accountable, and leave evidence of their actions?
a parallel locus shift: the hard problem moves from capability to the arrangements around it
-
Can regulation keep pace with AI's rapid evolution?
Current regulatory frameworks in the EU, US, and UK struggle to address generative AI's harms because rules become obsolete before they take effect. The question is whether dynamic regulation—one that adapts as quickly as models advance—is actually achievable.
the pacing frame this discussion sets aside
Related papers in this collection 8
Papers most semantically related to this note, ranked by cosine similarity in the embedding space.
- The Law of Stop: Interruptibility, Injunctions, and the Governance of Agentic AI
- Open-World Evaluations for Measuring Frontier AI Capabilities
- PACT: Can Enterprise AI Assistants Be Trusted Under Pressure?
- AI Agents Push Humans Out of the Loop
- Exploring Autonomous Agents: A Closer Look at Why They Fail When Completing Tasks
- Prompting Against Persona Drift: Comparing Intervention Timing and Content in LLM-Simulated Conversations
- AutoLab: Can Frontier Models Solve Long-Horizon Auto Research and Engineering Tasks?
- Norms at a Price: Why RL-Based Alignment Can Promise Conditional Compliance at Best
Original note title
measures that slow the frontier govern the conditions under which capabilities advance but do not resolve who may intervene when a deployed system causes harm or how