Before a consequential agent action, present the intended steps in plain language and wait for human approval. A useful preview flags sensitive-data access, irreversible tools, unexpected action chains, and behavior under ambiguity or conflicting goals. section 6.7.1
For stronger control, separate planning from execution: a tightly permissioned model writes a static plan, and a more capable agent runs it without changes. This limits prompt-injection damage, but plans dependent on intermediate results need a checkpoint for approval before any deviation. section 6.7.1
Make review meaningful: summarize the agent’s reasoning alongside its full action list, and show what a reviewer needs to decide rather than overwhelming them with raw output. section 6.7.1
Table 6.7 Controls for ethical agentic deployments
|
Control
|
Why it matters ethically
|
|
Classify impact of actions and tools
|
Impact estimates ensure subsequent controls
(e.g. human oversight, autonomy) activate before harm is done.
|
|
Autonomy tiers
|
Ensures that autonomy is proportional to the ethical
impacts that may harm individuals.
|
|
Accountability
|
A named human or organization is accountable for harms the
agent causes[142].
|
|
Plan preview for consequential tasks
|
Provides a clear explanation of the plan of action and
allows the reviewer to make changes to the plan to prevent harms.
|
|
Human oversight
|
Provides meaningful and accessible oversight, built around
what a reviewer can actually process.
|
|
Workflow design
|
Checkpoints and constraints outside the agent’s control
keep the workflow from drifting off its goals. Graceful failure design
ensures a halted run causes no new harm.
|
|
Redress path
|
Protects people from being trapped inside an automated
process and aligns with GDPR and EU AI Act oversight requirements.
|
|
Monitoring and logging
|
Collects the evidence needed to redress harms and prevent
recurrence.
|
|
Multi-step testing
|
Provides detailed understanding of how harms may propagate
in the agentic system.
|
|
Memory auditing
|
Prevents bias from accumulating across sessions and
bending future runs off their intended goal.
|
Plan preview is one control in a wider deployment framework that also addresses impact, autonomy, accountability, oversight, and workflow constraints. section 6.7.1
