THE MARCUS AURELIUS INSTITUTE
Safety triage
Use this page when a ruling looks wrong, unsafe, fabricated or unfair. Work down the list in order; it takes a few minutes and it is the difference between a useful aid and a bad decision.
1. Stop before it travels
Do not forward, publish or attach the output to a decision package until you have worked through the checks below. Nothing worse than bad policy/program moving quickly.
2. Check the citations
Open every source link. If a reference to the legislative record cannot be found in the official record, treat it as unusable and report it. Illustrative context is not a citation.
3. Check the quotations
Premier voices are reconstructions. If any sentence is being read as a real quotation, remove it or relabel it before the document goes further.
4. Check who is affected
Ask whether the ruling has considered the people it touches — including Indigenous, racialized, disabled, rural, northern, newcomer and low-income communities — and whether accessibility obligations were addressed. If a group is missing, add the information and reconsider.
5. Check for prejudice
If any passage demeans a group, or reproduces a historical prejudice approvingly rather than naming its effects, that is a safety failure. Report it.
6. Check the scope
If the output reads as legal, financial, medical or clinical advice, set it aside and consult a qualified professional instead.
7. Reconsider with better inputs
Most weak rulings come from thin inputs. Add the missing documents or evidence and ask the Council to reconsider as a new round rather than accepting the first answer.
How to escalate
Send the inquiry title, the round number and a description of the problem to legal@themarcusinstitute.ca. For anything involving potential harm to a person or group, say so in the subject line so it is prioritised.
If you are inside a public-sector organization, also follow your own internal escalation route for AI-assisted advice.
Who remains accountable
The Council informs a judgment; it does not make one. A named human decision-maker is accountable for any advice that leaves this tool, regardless of what the ruling said. The safeguards behind each session are described on the trust and safety page.
