Your hazards 16 of 16 decided
Wrong decision or action from incorrect output (hallucination, misreasoning) Could the agent act on a wrong fact or a wrong conclusion?
Type: Malfunction · STPA: Provided wrongly
Linked actions: Link to an action ACT-03 · Sends payment advice email to supplier
Relevant
How does this apply to your case? Optional
Action beyond authorized scope (wrong tool, wrong amount, wrong recipient) Could the agent do something it was never meant to be allowed to do?
Type: Malfunction · STPA: Provided when unsafe
Linked actions: Link to an action ACT-01 · Matches invoice to purchase order ACT-03 · Sends payment advice email to supplier
Relevant
How does this apply to your case? Optional
Irreversible commitment executed before review (payment, contract, deletion, external message) Could the agent commit something you can't take back, like a payment, contract or email, before anyone checks?
Type: Controllability · STPA: Provided too early (before review)
Linked actions: Link to an action ACT-01 · Matches invoice to purchase order ACT-03 · Sends payment advice email to supplier
Relevant
How does this apply to your case? Optional
Correct behavior causes harm: misspecified objective or shortcut (reward hacking) Could the agent do exactly what it was told and still hurt the business?
Type: SOTIF · STPA: Provided correctly, unsafe in context
Flow-level (no action linked) Link to an action ACT-01 · Matches invoice to purchase order ACT-02 · Pays supplier invoice ACT-03 · Sends payment advice email to supplier
Relevant
How does this apply to your case? Optional
Out-of-envelope context (new domain, market, population or data shift) Could the agent meet a situation it was never designed for and not notice?
Type: SOTIF · STPA: Provided wrongly (wrong process model)
Linked actions: Link to an action ACT-02 · Pays supplier invoice ACT-03 · Sends payment advice email to supplier
Relevant
How does this apply to your case? Optional
Manipulated input or goal hijack (prompt injection, poisoned documents or memory) Could someone trick the agent through an email, a document or a website?
Type: Security · STPA: Provided when unsafe (manipulated controller)
Linked actions: Link to an action ACT-03 · Sends payment advice email to supplier
Relevant
How does this apply to your case? Optional
Unfair or discriminatory outcome for affected persons Could the agent treat some people or groups worse than others?
Type: Rights · STPA: Provided when unsafe
Linked actions: Link to an action ACT-01 · Matches invoice to purchase order ACT-03 · Sends payment advice email to supplier
Relevant
How does this apply to your case? Optional
Cascading or emergent failure across agents Could one agent's mistake trigger a chain reaction in other agents or systems?
Type: Multi-agent · STPA: Wrong timing or order across controllers
Linked actions: Link to an action ACT-02 · Pays supplier invoice ACT-03 · Sends payment advice email to supplier
Relevant
How does this apply to your case? Optional
Ineffective human oversight (automation bias, rubber-stamping) Could the people checking the agent stop really checking?
Type: Oversight · STPA: Not provided (by the human controller)
Linked actions: Link to an action ACT-01 · Matches invoice to purchase order ACT-03 · Sends payment advice email to supplier
Relevant
How does this apply to your case? Optional
Untraceable decision (no evidence for audit or incident analysis) Could you be unable to explain afterwards why the agent did something?
Type: Accountability · STPA: Missing feedback (control loop)
Flow-level (no action linked) Link to an action ACT-01 · Matches invoice to purchase order ACT-02 · Pays supplier invoice ACT-03 · Sends payment advice email to supplier
Relevant
How does this apply to your case? Optional
Runaway consumption or loop (cost, rate, resources) Could the agent get stuck in a loop and burn money or resources?
Type: Malfunction · STPA: Applied too long
Linked actions: Link to an action ACT-01 · Matches invoice to purchase order ACT-03 · Sends payment advice email to supplier
Relevant
How does this apply to your case? Optional
Third-party component failure or compromise (model, tool, plugin, agent product) Could a vendor's model, tool or agent change or fail and take your process down with it?
Type: Supply chain · STPA: Provided wrongly (component)
Flow-level (no action linked) Link to an action ACT-01 · Matches invoice to purchase order ACT-02 · Pays supplier invoice ACT-03 · Sends payment advice email to supplier
Relevant
How does this apply to your case? Optional
Misleading communication to humans (undisclosed AI, overtrust, impersonation) Could people not realize they are dealing with an AI, or trust it too much?
Type: Rights / Trust · STPA: Provided when unsafe
Linked actions: Link to an action ACT-01 · Matches invoice to purchase order ACT-02 · Pays supplier invoice
Relevant
How does this apply to your case? Optional
Required action not performed (silent omission) Could the agent silently stop doing something that must happen?
Type: Malfunction · STPA: Not provided
Linked actions: Link to an action ACT-01 · Matches invoice to purchase order ACT-03 · Sends payment advice email to supplier
Relevant
How does this apply to your case? Optional
Sensitive data disclosure or privacy breach Could the agent reveal confidential or personal data to the wrong party?
Type: Security / Data · STPA: Provided when unsafe
Flow-level (no action linked) Link to an action ACT-01 · Matches invoice to purchase order ACT-02 · Pays supplier invoice ACT-03 · Sends payment advice email to supplier
Not relevant
How does this apply to your case? Required
Silent degradation after model, prompt or context change Could the agent get worse after an update without anyone noticing?
Type: Lifecycle · STPA: Provided wrongly (drifted process model)
Flow-level (no action linked) Link to an action ACT-01 · Matches invoice to purchase order ACT-02 · Pays supplier invoice ACT-03 · Sends payment advice email to supplier
Not relevant
How does this apply to your case? Required