{
  "id": 18453427,
  "title": "Singapore authorities warn of privacy, security risks from personal AI agents",
  "url": "https://urgent.news/2026/11/06/singapore-authorities-warn-of-privacy-security-risks-from-personal-ai-18453427",
  "topic": "ai",
  "section": "AI",
  "published": "2026-11-06T05:30:00.000Z",
  "source": {
    "name": "CNA - Singapore",
    "slug": "cna-singapore",
    "url": "https://www.channelnewsasia.com/singapore/personal-ai-agents-assistant-advisory-muse-dots-hermes-grok-openclaw-6529836"
  },
  "original_language": "en",
  "account": "Singapore's cybersecurity officials have issued a cautionary note about the risks associated with personal artificial intelligence (AI) agents. These AI agents, capable of managing emails, making bookings and accessing websites, can present privacy and security concerns for users. Unlike typical AI chatbots, personal AI agents can execute tasks on behalf of users, making them a potential target for scammers.\n\nAuthorities advise users to exercise caution when granting personal AI agents access to online accounts. Once connected, these agents can retrieve, process, and share information from various services, posing a risk of revealing sensitive personal data or enabling unauthorized actions. They may also carry out unintended tasks or be manipulated by malicious actors to perform harmful actions in the background, even when users are not actively using them.\n\nTo mitigate these risks, users are encouraged to limit the permissions granted to AI agents, retain control over crucial actions, and understand how to revoke access when necessary. Connecting only the required apps and accounts for an agent to perform its tasks, restricting the information it can access, and avoiding sharing sensitive credentials are also recommended. For services utilizing Singpass, users should opt for passkey authentication and avoid approving unexpected requests.\n\nSome AI agents may be able to send messages, make purchases, or perform other actions on behalf of users, depending on the permissions granted. Consumers are advised to start with limited or read-only permissions, require agents to seek approval for significant actions, and opt for one-time approvals instead of continuous authorization. Extra caution should be exercised when using AI agents for high-risk transactions, with users checking details before approving any action.\n\nThe authorities also warn consumers to review and revoke unnecessary permissions, monitor agents' activities, and understand how AI agent providers collect, store, and use personal information. Many personal AI agents have persistent memory, creating profiles of users and their contacts based on past interactions and files. Users should regularly review this information, delete inaccurate details, and ensure their information is not being used to train AI models or personalize advertisements.",
  "summary": "\"AI agents can be very useful, but you cannot treat them like human assistants and give them the same broad access to your information and implicit trust,\" said the authorities.",
  "key_points": [],
  "editors_take": null,
  "illustration": null,
  "coverage": {
    "outlets": 2,
    "also_reported_by": [
      {
        "outlet": "Straits Times",
        "title": "Singapore authorities urge strict limits on AI agent autonomy, access to personal accounts",
        "url": "https://urgent.news/2026/11/06/singapore-authorities-urge-strict-limits-on-ai-agent-autonomy-access",
        "published": "2026-11-06T02:31:21.000Z"
      }
    ]
  },
  "ai_generated": true,
  "disclaimer": "Summaries, key points and the editor’s take are written by software from other outlets’ reporting and may contain errors — always check the linked original."
}