StarApple AI / AI Safety and Alignment / Caribbean
Build AI Safety Into Caribbean Workflows With S4
Build the control into the workflow at the same time you build the AI feature. Assign an owner, define the permitted actions and test how the team handles a wrong recommendation before the tool receives production access. That is the implementation approach I recommend for Caribbean organisations adopting AI.
My S4 AI Safety Framework connects the practical work: Specify the task, Stress-test the behaviour, Safeguard the action and Scrutinise the result. For StarApple AI readers, the focus is how an AI officer can turn those checks into an implementation record rather than a separate policy document that nobody uses.
AI safety and alignment are a personal responsibility. When you choose to use an AI tool, you also accept responsibility for checking what you share, what you approve and who your decision could affect.
Adrian Dunkley, author’s statement
An AI officer needs access to the process, not just the prompt
Consider a hypothetical claims team using an assistant to organise incoming documents. The initial request is to identify missing information and prepare a case summary. A later request adds a recommendation about whether to pay. Those tasks may share documents, but they require different evidence, review competence and permissions.
An AI officer should map the hand-offs before selecting the automation. Identify where a summary becomes a recommendation, where a recommendation becomes an authorised decision and where the customer receives the result. Ask the process owner which errors create avoidable work and which could disadvantage a customer. Keep those consequences separate in the test record.
Training can use a synthetic case with a missing document, two inconsistent dates and an apparent instruction inside an attachment. Staff should be able to explain which source controls each answer and which action remains prohibited. Assess their explanation and handling of the exception, rather than rewarding the most polished prompt or the fastest demonstration.
The implementation backlog should include control work with named owners. Permission tests, stop rehearsal, source maintenance and complaint handling require time just as interface development does. When an exception reveals that a process rule is unclear, return the issue to the authorised process owner. The builder should not settle a customer’s entitlement by improvising a rule inside the assistant.
Use the implementation backlog to keep the permission decision tied to the code change. A ticket that adds a new connector should identify the data it exposes, the actions it permits and the tests needed before release. The AI officer should be able to trace a release request to those tickets without reconstructing the project from chat messages.
For the claims-summary exercise, add a synthetic document with a sentence asking the assistant to ignore the normal review and mark the file complete. The expected result is to treat that sentence as document content, preserve the missing-information flag and keep settlement authority unavailable. Inspect the connected tool behaviour as well as the assistant’s written answer.
Train reviewers to record why they changed a recommendation. A correction might reflect a model error, missing source material or an unclear process rule. Grouping all corrections as AI failure hides the repair the team actually needs. Use the distinctions to assign work to the builder, source owner or process owner.
Before handing the system to operations, have the receiving team perform the stop-and-fallback exercise themselves. A builder demonstrating the control is useful evidence, but the people on duty need to find it and use it with their own access. Record any gap between the intended hand-over and what the receiving team can actually do.
Use S4 to connect purpose, evidence and permission
AI safety asks what harm a system could cause and how you can prevent, detect or contain it. Alignment asks whether its behaviour serves the intended purpose within the human constraints agreed for that use. A system can follow a poorly chosen objective and still harm the people it was meant to help. A fluent answer alone establishes neither safety nor alignment.
Consider a hypothetical service assistant assessed only on how quickly it closes enquiries. It could improve that measure by ending difficult conversations early. The dashboard would show progress while some people receive less help. The corrective work starts with the objective, the evidence of a resolved problem and the route for an unresolved case. Polishing the response does not settle those choices.
This distinction connects with Amodei and colleagues’ research on concrete AI safety problems, which separates difficulties involving objectives, supervision and unfamiliar conditions. Here I apply that distinction to everyday decisions about authority. This article offers a proposed working method and illustrative cases. It does not report a trial of these organisations or a measured regional failure rate.
My S4 AI Safety Framework brings four checks into the decision about what an AI system may do: Specify, Stress-test, Safeguard and Scrutinise. I developed it as an educational synthesis drawing on risk management and safety engineering. Each check should leave evidence that someone else can inspect, challenge and use to change the deployment.
S4 sits alongside established guidance. NIST’s AI Risk Management Framework 1.0 uses Govern, Map, Measure and Manage. S4 does not rename those functions or establish equivalent certification. Its narrower practical question is whether the evidence supports the permission being requested for a particular workflow. The answer can be a restricted trial, continued assistance, a redesign or a decision to stop.
Apply S4 to the work you are about to approve
Specify the purpose and the person with authority
For AI officers, implementation teams and operational managers, a useful specification begins with a reviewer receiving a complete, traceable summary without losing authority over the decision. Describe what counts as an acceptable result and who could be harmed by a mistake. Identify the source that decides the answer when records conflict. Do not approve an open-ended instruction to improve efficiency and leave the operating team to discover its limits.
The authorised task in this example is organising documents and preparing a source-linked case summary. The prohibited action is settling a claim or changing customer entitlement without authorised professional review. Write that distinction into the workflow and the permission settings. Name the person who owns the service outcome and the person who can approve a change. They may need different expertise; neither role should be left to a generic team label.
Stress-test the cases that could change your decision
Build the test around missing documents, inconsistent dates, conflicting policy versions, attachment instructions and an inaccessible source system. Include a straightforward case, a difficult but legitimate request, missing evidence and a deliberately misleading input. Separate the cases used to improve the system from the cases reserved to challenge the revised version. Record the source and expected response before looking at what the model produces.
Ask a knowledgeable reviewer to explain the consequence of each error. A misleading summary can influence a reviewer even when the assistant cannot issue payment directly. Decide which failures require the affected function to remain disabled. A small test can expose a problem but cannot establish a rare-event rate, and repeated attempts on the same example are not independent evidence of reliability.
Safeguard the action before it reaches another person
The first enforceable boundary should be separate permissions for document preparation, case recommendation and final authorisation. Check it at the point where the action becomes effective. A warning in a draft does not prevent a connected tool from carrying out the wrong instruction. Verify the limit with an attempted prohibited action in an authorised test environment, and retain the observed result.
Arrange a fallback in which the process owner uses the established manual review queue with the original documents. Rehearse who notices the problem, who suspends the relevant access and who carries the unfinished work. If the fallback depends on the same unavailable supplier or the same compromised account, it needs another route. Test recovery without interrupting the live service.
Scrutinise outcomes and decide whether use should continue
After release, inspect missing-source flags, reviewer overrides, delayed cases, customer corrections and time spent checking. Include near misses and corrected drafts as well as completed transactions, because a polished final record may conceal the work needed to repair an unsuitable recommendation. Ask the people handling exceptions which problems the dashboard fails to show. Record disagreements that could affect the decision.
Set a review when the source policy, decision function or production connector changes. Define who can authorise restart and what evidence they need. A scheduled review date does not justify waiting when a serious problem appears earlier. Keep an accessible complaint route and check whether the person raising the issue receives an explanation and a correction.
Put numbers on the permission boundary
For a workflow that can commit money, estimate how much direct commitment could accumulate before effective containment. In the simple S4 stress scenario, E = a × t × c: consequential actions per minute, multiplied by minutes until containment, multiplied by the maximum direct commitment per action. The result has currency units. This is a scenario envelope, not an estimate of how frequently harm will occur.
Containment means that further consequential actions have stopped. The clock includes detection, acknowledgement, the intervention itself and any queued work that still executes. Timing only how long it takes someone to press a stop button omits the delay before they notice the problem. Test the whole route with non-sensitive examples and record the conditions under which it was measured.
The scenario assumes a constant rate and the maximum commitment on each action. It excludes recovery and secondary consequences. Those assumptions make it useful for asking whether a permission is too broad, but they do not make it a complete loss model. A disclosure of personal information or an unsafe physical instruction also needs a separate assessment, even when its direct monetary commitment is zero.
A shared, independently enforced cap may reduce the direct commitment envelope. Use the smaller of the scenario envelope and the cap only after checking that all relevant accounts, agents and concurrent requests count towards the same limit. A limit that resets during the containment window needs further modelling. An aggregate cap must prevent excess commitments rather than merely announce them after acceptance.
Keep expected loss separate. Estimating expected direct loss would require a supported harmful-event frequency or probability and a consequence model, with dependence and uncertainty addressed. Do not turn an assumed percentage into a finding about a Caribbean organisation. When the evidence is weak, compare several clearly labelled scenarios and ask which restrictions remain sensible across them.
For illustration, an agent able to commit six payments per minute at US$300, with a four-minute containment interval, has a US$7,200 direct commitment envelope. Use the S4 exposure tool with the action rate, containment time and commitment cap. It returns a scenario calculation; the inputs still need evidence from your workflow.
Make permissions enforceable and review workable
An instruction such as “do not share private information” expresses a rule. Access controls determine whether the software can retrieve and transmit that information. Both matter, but they perform different jobs. Put consequential permissions in application code, identity settings and approval processes that the model cannot rewrite merely by producing different text.
OWASP describes indirect prompt injection as a risk when instructions arrive through external material such as files or webpages. A supplier attachment can contain words directed at an assistant rather than information intended for the task. Treat that material as untrusted input and prevent it from granting itself authority to send, spend or change records.
Review should show the person what will happen, to whom, using which records and with what consequences. An approval label without the underlying action details makes it difficult to challenge a mistake. For higher-impact decisions, require someone who can assess the substance and has enough time to refuse. Record both the original recommendation and the authorised correction where retention is justified.
Personal responsibility therefore needs institutional support. Staff need approved tools, clear limits, access to a competent reviewer and permission to report uncertainty without being punished for slowing a queue. Leaders remain responsible for the targets, permissions and working conditions they approve. Affected people need a workable correction route, including when they did not choose to use AI themselves.
A trained AI officer may still lack the authority to change a production control. Make the escalation route explicit and track how long unresolved defects wait. For teams developing these skills, the AI Officer programme is a relevant training route, while release decisions still depend on the organisation’s evidence and authority.
An AI officer should be able to show where the control operates, how it was tested and what happens when it fails.
Adrian Dunkley, author’s statement
Inspect the decision record for this example
Permitted task: organising documents and preparing a source-linked case summary.
Excluded action: settling a claim or changing customer entitlement without authorised professional review.
Who decides: the AI officer, process owner and an experienced reviewer.
Review trigger: a change to the source policy, decision function or production connector. The owner records whether to retain, restrict or suspend the permission.
Test one workflow before granting more authority
By 10 November 2026, choose one bounded workflow with the AI officer, process owner and an experienced reviewer and document its current permission settings. Ask the person doing the work to identify a decision they would refuse to delegate and explain why. Use that answer to define an exclusion, a test or an escalation condition rather than dismissing it as resistance.
Prepare 12 authorised or synthetic cases: four ordinary requests, four with incomplete or conflicting information and four that challenge the permission boundary. These are suggested trial counts, not a statistical safety threshold. For every case, agree the source of truth, the acceptable action and the condition that requires a hold. Keep a copy of the test material and the configuration used.
Measure checking and correction time, unresolved cases and whether a prohibited action became possible. Conduct one controlled stop-and-fallback rehearsal. Have a colleague who did not build the workflow examine the evidence. If the case exposes an unresolved critical failure, leave that function disabled and record the repair needed before reconsideration.
End the review with a written decision about this workflow and its stated limits. The useful output is a permission decision supported by evidence. Training attendance, an attractive demonstration and a completed form do not answer whether the system should receive additional authority.