Cover of The Last Human Decision by Anzar Hasan
    Book companion page

    The Last Human Decision: Reader Companion

    Who Controls AI When Machines Can Think, Decide, and Act?

    by Anzar Hasan

    Available now on Amazon

    Tools and reference material from the book for boards, executives, and AI governance teams.

    Buy on Amazon

    Partner referral link — using it helps support OnAirToday. As an Amazon Associate I earn from qualifying purchases.

    Formats planned: Paperback (6 x 9 in, about 170 pages)

    The governing principle

    Govern the Access. Bound the Decision. Control the Action.

    1. Govern the Access

      Decide what the system can see, retrieve and invoke. Every AI system and agent gets its own identity and only the data, tools and permissions its approved purpose needs, so its reach is a deliberate choice rather than something it inherits.

    2. Bound the Decision

      Decide which choices the system may make alone, which need human approval, and which it must refuse. Writing these boundaries down before deployment turns autonomy into a governed grant of authority instead of a default.

    3. Control the Action

      Enforce limits at the moment of execution, independently of the model. Approval thresholds, prohibitions and rate limits are checked outside the model, so a persuasive or mistaken output cannot carry the system past them.

    The seven boardroom questions

    Ask these of every AI system before it is allowed to act.

    1. What decision does the system influence?
    2. What may it execute without approval?
    3. Which data and tools can it access?
    4. Which limits are enforced?
    5. Who owns the outcome?
    6. Who can intervene and how quickly?
    7. What evidence demonstrates that these controls work?

    Executive AI Decision Authority Checklist

    Work through it for one AI system at a time.

    0 of 18 complete

    Ownership and purpose
    Access
    Decisions
    Actions
    Assurance
    Change

    Your ticks are saved in this browser only.

    Frameworks referenced in the book

    The book presents these as resources to consider. It does not prescribe any standard.

    • ISO/IEC 42001
    • ISO/IEC 23894
    • NIST AI RMF 1.0 and the Generative AI Profile (AI 600-1)
    • NIST Cybersecurity Framework 2.0
    • EU AI Act
    • OWASP GenAI Security Project
    • MITRE ATLAS

    Downloads

    Printable PDF versions of these tools will be added here.

    Anzar Hasan

    Anzar Hasan

    Author, The Last Human Decision

    Scenarios in the book are hypothetical. This page is educational and is not legal, financial, cybersecurity or medical advice.