Saved in:
| Main Authors: | , , , , , , , , |
|---|---|
| Format: | Preprint |
| Published: |
2025
|
| Subjects: | |
| Online Access: | https://arxiv.org/abs/2504.12170 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1866917986940485632 |
|---|---|
| author | Stix, Charlotte Pistillo, Matteo Sastry, Girish Hobbhahn, Marius Ortega, Alejandro Balesni, Mikita Hallensleben, Annika Goldowsky-Dill, Nix Sharkey, Lee |
| author_facet | Stix, Charlotte Pistillo, Matteo Sastry, Girish Hobbhahn, Marius Ortega, Alejandro Balesni, Mikita Hallensleben, Annika Goldowsky-Dill, Nix Sharkey, Lee |
| contents | The most advanced future AI systems will first be deployed inside the frontier AI companies developing them. According to these companies and independent experts, AI systems may reach or even surpass human intelligence and capabilities by 2030. Internal deployment is, therefore, a key source of benefits and risks from frontier AI systems. Despite this, the governance of the internal deployment of highly advanced frontier AI systems appears absent. This report aims to address this absence by priming a conversation around the governance of internal deployment. It presents a conceptualization of internal deployment, learnings from other sectors, reviews of existing legal frameworks and their applicability, and illustrative examples of the type of scenarios we are most concerned about. Specifically, it discusses the risks correlated to the loss of control via the internal application of a misaligned AI system to the AI research and development pipeline, and unconstrained and undetected power concentration behind closed doors. The report culminates with a small number of targeted recommendations that provide a first blueprint for the governance of internal deployment. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2504_12170 |
| institution | arXiv |
| publishDate | 2025 |
| record_format | arxiv |
| spellingShingle | AI Behind Closed Doors: a Primer on The Governance of Internal Deployment Stix, Charlotte Pistillo, Matteo Sastry, Girish Hobbhahn, Marius Ortega, Alejandro Balesni, Mikita Hallensleben, Annika Goldowsky-Dill, Nix Sharkey, Lee Computers and Society The most advanced future AI systems will first be deployed inside the frontier AI companies developing them. According to these companies and independent experts, AI systems may reach or even surpass human intelligence and capabilities by 2030. Internal deployment is, therefore, a key source of benefits and risks from frontier AI systems. Despite this, the governance of the internal deployment of highly advanced frontier AI systems appears absent. This report aims to address this absence by priming a conversation around the governance of internal deployment. It presents a conceptualization of internal deployment, learnings from other sectors, reviews of existing legal frameworks and their applicability, and illustrative examples of the type of scenarios we are most concerned about. Specifically, it discusses the risks correlated to the loss of control via the internal application of a misaligned AI system to the AI research and development pipeline, and unconstrained and undetected power concentration behind closed doors. The report culminates with a small number of targeted recommendations that provide a first blueprint for the governance of internal deployment. |
| title | AI Behind Closed Doors: a Primer on The Governance of Internal Deployment |
| topic | Computers and Society |
| url | https://arxiv.org/abs/2504.12170 |