In development
AI Agent Certification
Who is answerable when an AI agent gets it wrong?
We attest that an AI agent follows the rules, doesn't make things up and doesn't deceive, before the public uses it.
A tribunal ordered Air Canada to pay for what its chatbot promised.Source: Moffatt v. Air Canada, 2024 BCCRT 149 (February 14, 2024)
What we check
Three things, before the public uses it.
It follows the rules
That the agent respects the rules and laws that apply to it, and its own instructions.
It doesn't make things up
That what it states can be checked against a source, rather than coming out of nowhere.
It doesn't deceive
That it doesn't promise what it can't deliver or pretend to be something it isn't.
Why it matters
Accountability is no longer optional.
A tribunal has already ruledA tribunal ordered Air Canada to pay for what its chatbot promised. Source: canlii.org — Moffatt v. Air Canada, 2024 BCCRT 149
EU AI ActHigh-risk AI systems need human oversight, record-keeping and transparency. Their obligations apply from December 2, 2027 (Regulation (EU) 2024/1689). Source: eur-lex.europa.eu — Regulation (EU) 2024/1689
How we see it
Before it reaches the public.
- 01You tell us what your agent doesAnd under which rules it must work.
- 02We put it to the testBefore any person uses it.
- 03We leave a recordOf what it meets and of what needs fixing.
- 04A person decidesHuman oversight always has the last word.
Are you about to put an AI agent in front of the public?
In development. Tell us what it does and we'll talk.