RESPONSIBLE AI
Safety and accountability principles for AI use
Use AI with a clear purpose, minimum necessary data, verification, and human responsibility.
Do
- Define the purpose and expected output before using AI.
- Keep personal and sensitive data out of unauthorized public tools.
- Verify outputs before using them in payroll, legal, hiring, or employee decisions.
- Record human review for high-impact decisions.
- Report inaccurate, biased, or unsafe outputs.
Do not
- Turn an AI output into a hiring, promotion, pay, or termination decision without human review.
- Present unverified output as an official source or research finding.
- Hide uncertainty, limitations, or model errors.
Quick checklist
- Is AI necessary for this task?
- Is the data minimized?
- Who verifies the output?
- How will the process stop if bias or error is found?
- Can an affected person appeal?