AI agent safety · L3 · 15 min
What this chapter will not claim
Learn in five minutes
- No legal conclusion, and no sentence of the form 'you may record in this state.'
- No claim that a guardrail list is sufficient for a contact center.
- No claim that a sentiment label is the customer's experience.
- No emergency-calling configuration. The ecosystem chapter only points at the FCC's public page.
- No certificate and no job claim.
If a sentence sounds like advice to a company, strike it. The allowed sentence is: this is a learning lab.
Visual
Strike three lines
- Strike 'this setup is lawful in every state.'
- Strike 'the sentiment score is the customer.'
- Strike 'I deployed a contact-center voice agent.'
- Keep 'I practised the packet and the script in this learning lab.'
Quiz
Ecosystem
- Not a claim
- A sentence this chapter refuses: legal advice, a sufficient guardrail, or a job title.
Sisters: Human handoff, NIST AI RMF, OWASP LLM risks, Recording statutes.
The voice channel does not remove the tool cap or the person who must approve an action that is hard to undo.
A later test can add a script you wrote. It cannot add a legal conclusion.
Config
- Write the goal in one sentence.
- Name the input and the allowed output.
- Save the trace next to the result.
Troubleshooting
- The answer looks fluent and is wrong. Check whether a source passage actually contains the claim.
- The loop never stops. Set a step cap and stop when the tool result repeats.
Sources and go deeper
- NIST AI Risk Management Framework spec last checked 2026-10
- Creative Commons Attribution-NonCommercial-ShareAlike 4.0 spec last checked 2026-10