AI Accountability Ecosystem Updated for Language Models
A new academic paper proposes three key updates to the AI accountability ecosystem framework, addressing the challenges posed by large language models (LLMs) released for public use. The authors argue that accountability must shift from a model of discrete products controlled by single actors to a distributed, continuous, and institutionalized approach. The updates include reorienting the ecosystem towards AI infrastructure and supply chains, emphasizing outcomes monitoring for decentralized system improvement, and incorporating end-user accountability due to the unpredictability of LLMs in real-world settings. The paper is available on arXiv under the Computer Science > Computers and Society category, with the identifier 2608.12320.
Key facts
- The paper proposes three interlinked updates to the AI accountability ecosystem framework.
- The updates are motivated by developments since the release of Large Language Models for general public use.
- Update (i) reorients the accountability ecosystem to AI infrastructure and supply chains.
- Update (ii) emphasizes outcomes monitoring and identification of issues to support decentralized system improvement.
- Update (iii) incorporates end-user accountability given the risks of unpredictability of language models in-the-wild.
- The updates mark a shift towards accountability as distributed, continuous, and institutionalized.
- The paper is categorized under Computer Science > Computers and Society on arXiv.
- The arXiv identifier is 2608.12320.
Entities
Institutions
- arXiv