Illustrative examples of personal AI-agents
| # | Example | Relevance to personal AI-agents for CXV | Source(s) |
|---|---|---|---|
| 1 | EU Artificial Intelligence Act (2024) | A definition of AI systems adopted by EU regulations and an illustration of regulatory standards to govern personal AI-agents. By influencing and following these regulatory frameworks, the design attribute of “does not exploit vulnerability” can be potentially implemented | https://artificialintelligenceact.eu/the-act/ |
| 2 | Project VRM | An illustration of how inversion of control can be achieved by shifting the control of relationship from the service providers to consumers, especially for those experiencing vulnerability | https://projectvrm.org/ |
| 3 | Algorithmic Trading (IG) | An illustration where delegated authority can be managed, maintained and controlled by consumers to where AI-agents perform automated trading on behalf of consumers | https://www.ig.com/uk/trading-platforms/algorithmic-trading |
| 4 | Neuralink | An illustration of inversion of control in a personal AI-agent that can adapt to the brain function of a consumer to influence the perceived control in favor of the consumer experiencing visual vulnerabilities to more effectively interactions with digital service providers through computers. At the time of authoring, the project goal of the Neuralink implant's wireless brain-computer interface (BCI) design is to “restore autonomy to people with paralysis” | https://greekreporter.com/2024/09/22/elon-musk-neuralink-blindsight-approved-breakthrough-device-fda/ https://neuralink.com/updates/a-year-of-telepathy/ |
| 5 | Kwaai.ai | A reference implementation of the Project VRM (above) on how inversion of control can be provided as a service by AI vendors in the market | https://www.kwaai.ai |
| 6 | A2A Protocol (Agent to Agent protocol) | An illustration of protocols that promote interoperability standards. Personal AI-agents can adopt such standards to enhance interoperability | https://a2a-protocol.org/latest |
| 7 | Model Context Protocol | An illustration of how personal AI-agents can interact with service systems that may or may not offer an AI-agent interface. This example allows personal AI-agents to be more adaptable to vulnerability as well as interoperable to access external systems and processes | https://modelcontextprotocol.io/docs/getting-started/intro |
| 8 | Open Banking | An illustration of a reference implementation of interoperability standards set out by open banking. Consumers experiencing vulnerability have access to payment, transactions and banking information from multiple sources as well as seamlessly port from one provider to the other | https://www.openbanking.org.uk/ |
| 9 | Siri computer application | An illustration of a natural language interface that can orchestrate several services that are available on Apple devices such as map navigation, listening to music, notifications or plain conversations as a companion | https://www.britannica.com/technology/natural-language-processing-computer-science |
| 10 | Proton VPN “Privacy policy” | A simple illustration of the protective sentinel role implemented by a VPN provider in seamlessly securing connections from consumers experiencing vulnerability to provide control to consumers on what behaviors can be screened and monitored | https://proton.me/legal/privacy |
| 11 | Create Agent To Track Personal Finances with GPT-4 via Telegram and Google Sheets |N8n Workflow Template | A simple illustration of an “autonomous ally” that can be self-created. At the time of authoring this paper, several large language model providers, such as open AI's Agent kit are working on simplifying consumers to create AI-agents | https://n8n.io/workflows/3932-track-personal-finances-with-gpt-4-via-telegram-and-google-sheets https://openai.com/index/introducing-agentkit/ |
| 12 | Flow Neuroscience Headset | An illustration of adaptation to vulnerability in a Reliable intermediary role. The headset is an example of early manifestations of the future of embodied AI. Patients suffering from depression can be supported by a neutral and non-invasive intermediary in the form of a physical headset | https://news.sky.com/story/electric-headset-for-treating-depression-trialled-by-nhs-12884293 |
| 13 | UK Digital Identity and Attributes Trust Framework Recommendation of the Council on Artificial Intelligence | Illustrations of identity, trust and AI standards that policy makers can devise to support service providers as well as consumers experiencing vulnerability. Personal AI-agents can adhere to these standards by to implement “delegated authority” as a design attribute | https://www.gov.uk/government/collections/uk-digital-identity-and-attributes-trust-framework https://legalinstruments.oecd.org/en/instruments/OECD-LEGAL-0449 |
| 14 | Digital public infrastructure (DPI) United Nations Development Programme, Digital, AI and Innovation Hub | An illustration of how governments can setup the digital infrastructure for AI-agents to promote innovation and the personal AI-agent design attributes can be adopted by AI-agents that are implemented on such an infrastructure | htttps://www.undp.org/digital/digital-public-infrastructure |
| 15 | W3C Linked Web Storage Working Group Charter (Solid) | An illustration of interoperability protocols and standards that promote both interoperability and inversion of control | https://www.w3.org/2024/09/linked-web-storage-wg-charter.html |
| 16 | ISO standard for “Consumer vulnerability — requirements and guidelines for the design and delivery of inclusive service” | An illustration of standards the can be adopted by personal AI-agents, service providers and policy makers in understanding vulnerability requirements and needs | https://www.iso.org/standard/73261.html |
| # | Example | Relevance to personal AI-agents for CXV | Source(s) |
|---|---|---|---|
| 1 | A definition of AI systems adopted by | ||
| 2 | Project | An illustration of how inversion of control can be achieved by shifting the control of relationship from the service providers to consumers, especially for those experiencing vulnerability | |
| 3 | Algorithmic Trading ( | An illustration where delegated authority can be managed, maintained and controlled by consumers to where AI-agents perform automated trading on behalf of consumers | |
| 4 | Neuralink | An illustration of inversion of control in a personal AI-agent that can adapt to the brain function of a consumer to influence the perceived control in favor of the consumer experiencing visual vulnerabilities to more effectively interactions with digital service providers through computers. At the time of authoring, the project goal of the Neuralink implant's wireless brain-computer interface (BCI) design is to “restore autonomy to people with paralysis” | |
| 5 | Kwaai.ai | A reference implementation of the Project | |
| 6 | A2A Protocol (Agent to Agent protocol) | An illustration of protocols that promote interoperability standards. Personal AI-agents can adopt such standards to enhance interoperability | |
| 7 | Model Context Protocol | An illustration of how personal AI-agents can interact with service systems that may or may not offer an AI-agent interface. This example allows personal AI-agents to be more adaptable to vulnerability as well as interoperable to access external systems and processes | |
| 8 | Open Banking | An illustration of a reference implementation of interoperability standards set out by open banking. Consumers experiencing vulnerability have access to payment, transactions and banking information from multiple sources as well as seamlessly port from one provider to the other | |
| 9 | Siri computer application | An illustration of a natural language interface that can orchestrate several services that are available on Apple devices such as map navigation, listening to music, notifications or plain conversations as a companion | |
| 10 | Proton | A simple illustration of the protective sentinel role implemented by a | |
| 11 | Create Agent To Track Personal Finances with GPT-4 via Telegram and Google Sheets |N8n Workflow Template | A simple illustration of an “autonomous ally” that can be self-created. At the time of authoring this paper, several large language model providers, such as open AI's Agent kit are working on simplifying consumers to create AI-agents | |
| 12 | Flow Neuroscience Headset | An illustration of adaptation to vulnerability in a Reliable intermediary role. The headset is an example of early manifestations of the future of embodied AI. Patients suffering from depression can be supported by a neutral and non-invasive intermediary in the form of a physical headset | |
| 13 | UK Digital Identity and Attributes Trust Framework | Illustrations of identity, trust and AI standards that policy makers can devise to support service providers as well as consumers experiencing vulnerability. Personal AI-agents can adhere to these standards by to implement “delegated authority” as a design attribute | |
| 14 | Digital public infrastructure (DPI) United Nations Development Programme, Digital, AI and Innovation Hub | An illustration of how governments can setup the digital infrastructure for AI-agents to promote innovation and the personal AI-agent design attributes can be adopted by AI-agents that are implemented on such an infrastructure | |
| 15 | W3C Linked Web Storage Working Group Charter (Solid) | An illustration of interoperability protocols and standards that promote both interoperability and inversion of control | |
| 16 | An illustration of standards the can be adopted by personal AI-agents, service providers and policy makers in understanding vulnerability requirements and needs |