Amendment 13 is active. AI decisions are being reviewed. The question is no longer only whether AI was used — it is whether responsible human judgment can be shown in a reviewable record.
Boards hold personal accountability for data protection practices — including AI systems that process personal data. The Privacy Protection Authority has published draft guidance on AI-related obligations (April 2025) and is moving to active enforcement in 2026.
Israel Supreme Court · AAM 63194-08-25 · 22 March 2026: the court found a municipality acted «recklessly» by acting on AI output without documented human verification. The question was not whether AI was used — it was whether a human decision was on record.
Article 14 requires that high-risk AI systems be designed and deployed to allow meaningful human oversight — not passive monitoring, but the ability to understand, intervene, and correct. The EU extended the high-risk compliance timeline; the window to build this correctly is now. Article 50 (transparency requirements) applies from August 2026. High-risk oversight obligations under Article 14 apply from December 2027.
Effective July 30, 2026, Google's updated Terms of Service reinforce responsibility for AI-generated outputs by the organizations that act on them.
AI audit trails, model monitoring, and risk dashboards document what the AI did. They do not document what the human decided. A log entry that says «reviewed» is not evidence of meaningful review. A policy that says «humans oversee» is not a record of oversight. Regulators are beginning to draw this distinction.
Not a log of AI activity. Not a policy document. A record of the human: who reviewed, what they considered, what they decided — in a form designed to support later review by governance, regulatory, legal, or institutional stakeholders.
The record survives the system it was made in. It is not a screen capture. It is not a checkbox. It is a structured entry that documents the moment a human decision-maker was present — and decided.
Map the AI-assisted decisions in your workflow that carry regulatory, legal, or reputational weight — credit decisions, risk assessments, compliance determinations, medical or legal recommendations.
At each decision point, a structured prompt asks the accountable human to confirm review and record their judgment — not a checkbox, but a timestamped, structured entry that distinguishes meaningful engagement from passive sign-off.
The entry is anchored outside the AI system — in a form that survives the system it was made in, accessible to the board, auditors, and regulators on request.
When a regulator, auditor, or legal proceeding asks «who was the decision-maker, and when did they decide?» — the answer is a dated, structured record, not a policy document or a system log.
If your board is reviewing AI accountability obligations — or if a regulator, auditor, or legal proceeding has raised questions about human oversight — a brief, confidential conversation is the right next step.
Or write directly: hello@caneni.net
Currently a closed pilot · conversations confidential · one-page brief on request.
Full methodology: canon.caneni.net
← Back to Caneni
