# AI agents: what to check before they act Before an agent takes action, make its limits and approval points clear. The Assurance Desk. Published 8 October 2026. Last checked 8 October 2026. Part of the guide [What is AI governance, in plain words?](https://theassurancedesk.com/guides/what-is-ai-governance.md). **The short answer.** Set clear limits on what an AI agent can access and do, and decide where a person must approve its actions. Singapore's IMDA framework recommends these checks and keeps humans accountable. This is practical guidance from Singapore, not a statement of UK law ([imda.gov.sg, Singapore Launches New Model AI Governance Framework for Agentic AI, IMDA](https://www.imda.gov.sg/resources/press-releases-factsheets-and-speeches/press-releases/2026/new-model-ai-governance-framework-for-agentic-ai), checked 8 October 2026). ## What makes an AI agent different? An AI agent can work through a task and take action for its user. That might mean changing a customer record or making a payment. Its access to data and tools is therefore part of the decision about how to use it ([imda.gov.sg, Singapore Launches New Model AI Governance Framework for Agentic AI, IMDA](https://www.imda.gov.sg/resources/press-releases-factsheets-and-speeches/press-releases/2026/new-model-ai-governance-framework-for-agentic-ai), checked 8 October 2026). Oliver Patel highlights a practical question for organisations: how to stop agents accessing systems and sharing data in ways they should not. His observation puts the focus on the agent's reach ([Oliver Patel on LinkedIn](https://www.linkedin.com/posts/oliver-patel_mapping-official-frameworks-and-guidelines-activity-7506678366461227008-bmig), checked 8 October 2026). > "how to prevent agents from accessing systems and sharing data inappropriately" > Oliver Patel, LinkedIn post excerpt > Oliver Patel on LinkedIn, Agentic AI Governance Frameworks and Guidelines, Oliver Patel, AIGP, CIPP/E, MSc posted on the topic, https://www.linkedin.com/posts/oliver-patel_mapping-official-frameworks-and-guidelines-activity-7506678366461227008-bmig (checked 8 October 2026) > [View LinkedIn post](https://www.linkedin.com/posts/oliver-patel_mapping-official-frameworks-and-guidelines-activity-7506678366461227008-bmig) ## What should an agent be allowed to do? IMDA recommends choosing suitable tasks and limiting the agent's freedom to act. That includes which tools and data it can access. A useful starting question is: what does this agent need to do its job ([imda.gov.sg, Singapore Launches New Model AI Governance Framework for Agentic AI, IMDA](https://www.imda.gov.sg/resources/press-releases-factsheets-and-speeches/press-releases/2026/new-model-ai-governance-framework-for-agentic-ai), checked 8 October 2026)? The framework includes an example from Dayos, which handles IT support requests. Dayos assessed each type of request for its impact, whether an action could be undone, and whether a person could oversee it ([imda.gov.sg, Updated Model AI Governance Framework for Agentic AI, IMDA](https://www.imda.gov.sg/resources/press-releases-factsheets-and-speeches/factsheets/2026/updated-model-ai-governance-framework-for-agentic-ai), checked 8 October 2026). Dayos then used different levels of permission. Some actions can run automatically, others need approval, and some are off limits. The diagram shows that company's approach. It is an example to discuss, not a set of permissions to copy into every firm ([imda.gov.sg, Updated Model AI Governance Framework for Agentic AI, IMDA](https://www.imda.gov.sg/resources/press-releases-factsheets-and-speeches/factsheets/2026/updated-model-ai-governance-framework-for-agentic-ai), checked 8 October 2026). **Dayos's example: different actions, different permissions** - **Can act automatically** - Low impact actions that can be fully undone - Actions are checked afterwards - **Must ask first** - Moderate impact actions that can only partly be undone - The agent can propose a fix but needs human approval to carry it out - **Cannot act** - High impact actions that are hard to undo - Includes changing permissions in this example These are alternatives in one IT support example in IMDA's guidance. The agent's permission depends on the action's impact and whether it can be undone. Source: imda.gov.sg, Updated Model AI Governance Framework for Agentic AI, IMDA, https://www.imda.gov.sg/resources/press-releases-factsheets-and-speeches/factsheets/2026/updated-model-ai-governance-framework-for-agentic-ai (checked 8 October 2026) ## What does useful human approval look like? Useful human approval starts with clear checkpoints, as IMDA recommends. The guidance also describes automation bias. That means trusting a system too much because it has worked well before ([imda.gov.sg, Singapore Launches New Model AI Governance Framework for Agentic AI, IMDA](https://www.imda.gov.sg/resources/press-releases-factsheets-and-speeches/press-releases/2026/new-model-ai-governance-framework-for-agentic-ai), checked 8 October 2026). In IMDA's Tencent example, actions such as editing files need human approval by default. Complex computer commands are explained in plain English to help the user decide. Suspicious commands still need approval even if they were approved in advance ([imda.gov.sg, Updated Model AI Governance Framework for Agentic AI, IMDA](https://www.imda.gov.sg/resources/press-releases-factsheets-and-speeches/factsheets/2026/updated-model-ai-governance-framework-for-agentic-ai), checked 8 October 2026). For a buying or setup discussion, ask the team to show the approval screen and explain which actions trigger it. The Tencent example gives useful questions to ask ([imda.gov.sg, Updated Model AI Governance Framework for Agentic AI, IMDA](https://www.imda.gov.sg/resources/press-releases-factsheets-and-speeches/factsheets/2026/updated-model-ai-governance-framework-for-agentic-ai), checked 8 October 2026). **Questions to ask about human approval, drawn from IMDA's Tencent example** | Ask the team | What to look at together | | --- | --- | | Which actions need a person to approve them? | The default rules for file changes, computer commands and use of external tools | | Can the person understand what they are approving? | The plain English explanation shown before a complex command runs | | What happens if an approved command looks suspicious? | Whether the system asks for fresh approval | Source: imda.gov.sg, Updated Model AI Governance Framework for Agentic AI, IMDA, https://www.imda.gov.sg/resources/press-releases-factsheets-and-speeches/factsheets/2026/updated-model-ai-governance-framework-for-agentic-ai (checked 8 October 2026) ## How can you test before giving wider access? Test the agent and control which services it can reach, as IMDA recommends. Its GovTech Singapore example shows a small start before wider use ([imda.gov.sg, Singapore Launches New Model AI Governance Framework for Agentic AI, IMDA](https://www.imda.gov.sg/resources/press-releases-factsheets-and-speeches/press-releases/2026/new-model-ai-governance-framework-for-agentic-ai) and [imda.gov.sg, Updated Model AI Governance Framework for Agentic AI, IMDA](https://www.imda.gov.sg/resources/press-releases-factsheets-and-speeches/factsheets/2026/updated-model-ai-governance-framework-for-agentic-ai), checked 8 October 2026). GovTech Singapore first limited its coding tool to employees working on low risk systems, with no external tools. It prepared shared records of activity, monitoring and rules for connecting approved external tools. It also tested potential attacks to check the controls worked ([imda.gov.sg, Updated Model AI Governance Framework for Agentic AI, IMDA](https://www.imda.gov.sg/resources/press-releases-factsheets-and-speeches/factsheets/2026/updated-model-ai-governance-framework-for-agentic-ai), checked 8 October 2026). That example offers questions for a trial. Who can use the agent? Which systems can it reach? How will the team check its behaviour? It describes one rollout, not a promise that the same setup suits every firm ([imda.gov.sg, Updated Model AI Governance Framework for Agentic AI, IMDA](https://www.imda.gov.sg/resources/press-releases-factsheets-and-speeches/factsheets/2026/updated-model-ai-governance-framework-for-agentic-ai), checked 8 October 2026). **GovTech Singapore's example: a small start, then controls for wider use** - Limit the first use: Employees only, low risk systems and no external tools - Prepare for wider use: Shared activity records, monitoring and rules for approved tool connections - Test the controls: Check how the system responds to potential attacks These layers summarise the example in IMDA's guidance. They are not a test score or a legal checklist. Source: imda.gov.sg, Updated Model AI Governance Framework for Agentic AI, IMDA, https://www.imda.gov.sg/resources/press-releases-factsheets-and-speeches/factsheets/2026/updated-model-ai-governance-framework-for-agentic-ai (checked 8 October 2026) ## What should people using the agent know? People using an agent need clear information and training, under IMDA's approach. Its Workday example says users are told upfront that the tool uses AI ([imda.gov.sg, Singapore Launches New Model AI Governance Framework for Agentic AI, IMDA](https://www.imda.gov.sg/resources/press-releases-factsheets-and-speeches/press-releases/2026/new-model-ai-governance-framework-for-agentic-ai) and [imda.gov.sg, Updated Model AI Governance Framework for Agentic AI, IMDA](https://www.imda.gov.sg/resources/press-releases-factsheets-and-speeches/factsheets/2026/updated-model-ai-governance-framework-for-agentic-ai), checked 8 October 2026). Bring those points into the rollout discussion. How will people know they are using an AI agent? What training will they receive? How will approval points be explained ([imda.gov.sg, Singapore Launches New Model AI Governance Framework for Agentic AI, IMDA](https://www.imda.gov.sg/resources/press-releases-factsheets-and-speeches/press-releases/2026/new-model-ai-governance-framework-for-agentic-ai), checked 8 October 2026)? ## Does this apply if you buy the agent? Yes, the framework covers agents you buy as well as those you build. Its questions about access, human approval, testing and user training cover both routes ([imda.gov.sg, Singapore Launches New Model AI Governance Framework for Agentic AI, IMDA](https://www.imda.gov.sg/resources/press-releases-factsheets-and-speeches/press-releases/2026/new-model-ai-governance-framework-for-agentic-ai), checked 8 October 2026). Use those topics to structure a conversation with the supplier or your own team. Keep the country and purpose clear: this is Singapore's practical governance guidance. It is not proof that a tool meets UK law ([imda.gov.sg, Singapore Launches New Model AI Governance Framework for Agentic AI, IMDA](https://www.imda.gov.sg/resources/press-releases-factsheets-and-speeches/press-releases/2026/new-model-ai-governance-framework-for-agentic-ai), checked 8 October 2026). ## What to do 1. List the task, tools and data the agent needs, then discuss the limits on its access ([imda.gov.sg, Singapore Launches New Model AI Governance Framework for Agentic AI, IMDA](https://www.imda.gov.sg/resources/press-releases-factsheets-and-speeches/press-releases/2026/new-model-ai-governance-framework-for-agentic-ai), checked 8 October 2026). 2. Identify the points where a person must approve an action ([imda.gov.sg, Singapore Launches New Model AI Governance Framework for Agentic AI, IMDA](https://www.imda.gov.sg/resources/press-releases-factsheets-and-speeches/press-releases/2026/new-model-ai-governance-framework-for-agentic-ai), checked 8 October 2026). 3. Ask how the agent will be tested and kept within its approved services ([imda.gov.sg, Singapore Launches New Model AI Governance Framework for Agentic AI, IMDA](https://www.imda.gov.sg/resources/press-releases-factsheets-and-speeches/press-releases/2026/new-model-ai-governance-framework-for-agentic-ai), checked 8 October 2026). 4. Check how users will be told about the agent and trained to use it ([imda.gov.sg, Singapore Launches New Model AI Governance Framework for Agentic AI, IMDA](https://www.imda.gov.sg/resources/press-releases-factsheets-and-speeches/press-releases/2026/new-model-ai-governance-framework-for-agentic-ai), checked 8 October 2026). ## Common questions ### Must a person approve every AI agent action? IMDA recommends meaningful approval points. Its Dayos example allows some actions to run automatically, requires approval for others and blocks some altogether. Those are choices in that example, not universal settings ([imda.gov.sg, Singapore Launches New Model AI Governance Framework for Agentic AI, IMDA](https://www.imda.gov.sg/resources/press-releases-factsheets-and-speeches/press-releases/2026/new-model-ai-governance-framework-for-agentic-ai) and [imda.gov.sg, Updated Model AI Governance Framework for Agentic AI, IMDA](https://www.imda.gov.sg/resources/press-releases-factsheets-and-speeches/factsheets/2026/updated-model-ai-governance-framework-for-agentic-ai), checked 8 October 2026). ### Is this a UK legal checklist? No. The framework used here is Singapore's guidance on responsible agent use. This article explains its practical approach; it does not set out the laws that apply to a particular UK firm ([imda.gov.sg, Singapore Launches New Model AI Governance Framework for Agentic AI, IMDA](https://www.imda.gov.sg/resources/press-releases-factsheets-and-speeches/press-releases/2026/new-model-ai-governance-framework-for-agentic-ai), checked 8 October 2026). ### Who remains accountable when the agent acts? IMDA's framework emphasises that humans remain accountable. It recommends clear checkpoints where human approval is needed ([imda.gov.sg, Singapore Launches New Model AI Governance Framework for Agentic AI, IMDA](https://www.imda.gov.sg/resources/press-releases-factsheets-and-speeches/press-releases/2026/new-model-ai-governance-framework-for-agentic-ai), checked 8 October 2026). ## Where next - [What is AI governance, in plain words?](https://theassurancedesk.com/guides/what-is-ai-governance.md) - [What to ask an AI supplier before you buy](https://theassurancedesk.com/guides/what-to-ask-ai-supplier-before-you-buy.md) - [A DPIA for an AI tool: when and how](https://theassurancedesk.com/guides/dpia-for-an-ai-tool.md) ## Sources - imda.gov.sg, Singapore Launches New Model AI Governance Framework for Agentic AI, IMDA: https://www.imda.gov.sg/resources/press-releases-factsheets-and-speeches/press-releases/2026/new-model-ai-governance-framework-for-agentic-ai (checked 8 October 2026) - Oliver Patel on LinkedIn: https://www.linkedin.com/posts/oliver-patel_mapping-official-frameworks-and-guidelines-activity-7506678366461227008-bmig (checked 8 October 2026) - imda.gov.sg, Updated Model AI Governance Framework for Agentic AI, IMDA: https://www.imda.gov.sg/resources/press-releases-factsheets-and-speeches/factsheets/2026/updated-model-ai-governance-framework-for-agentic-ai (checked 8 October 2026) --- **Who publishes this.** The Assurance Desk is published by Add-Venture Advisory Ltd, company number 11854524, registered in England and Wales. The same owner runs Avouch, a marketing agency that sells AI visibility services to firms of the kind this site covers. That is a conflict of interest, and it is disclosed here.