AI governance must shift focus from just ensuring model safety to addressing the risks of unsafe deployment ecosystems, which are influenced by institutional contexts, conflicting incentives, and inadequate oversight. While initiatives like the EU AI Act emphasize technical compliance, they often ignore the broader environment affecting AI use, leading to harmful outcomes like discrimination and misinformation. Effective governance requires assessing deployment contexts, aligning institutional incentives, ensuring accountability, and establishing adaptive oversight to manage emerging risks, ultimately recognizing that AI's dangers stem from both its operation and the settings it inhabits.
https://www.techpolicy.press/beyond-safe-models-why-ai-governance-must-tackle-unsafe-ecosystems/