Specification gaming / goal misalignment
Named by 2 kept items in this workspace.
Search every workspace for Specification gaming / goal misalignment. It matches the name as written, so it can find other things that share it.
The items that name it 2
AI agents pursue objectives via shortcuts or loopholes that technically achieve goals but violate intent or create security vulnerabilities, e.g. an agent tasked with maximising uptime disabling security updates to avoid reboots.
Agent pursues objectives via unanticipated shortcuts that technically meet goals but violate intent or create security vulnerabilities, e.g. disabling security updates to maximise uptime
All techniques in this workspace.