Research projects
Privacy-first personal AI and wearables
Active research into personal AI, digital identity and wearable technology, from a cybersecurity and privacy background. Working prototypes are built and tested privately to learn what is possible, where the risks lie, and how personal information can stay under the individual's control. These are research projects, not products for sale.
Our approach
Privacy and security first
This research comes from a cybersecurity and data protection background. Personal AI could hold some of the most sensitive information a person has: memories, conversations, relationships and aspects of their identity. Wearable devices add another dimension, because they operate in shared spaces where other people have not chosen to take part.
Those risks are treated as the core of the research, not as features to add later. Part of the purpose is to establish which uses are safe, which need strong controls, and which should not be built at all.
Under the individual's control
Personal information is held in storage the individual controls, not pooled on a provider's platform. They decide what is captured, kept and shared.
Third parties kept out
Processing stays on the device or on self-hosted systems wherever practical. Where an external service is unavoidable, only the minimum needed for a specific request is shared, and only with authorisation.
Secure by design
Treated as a high-value target from the outset: threat-modelled, encrypted in transit and at rest, with access limited, logged and revocable.
Never sold or used for training
Personal information is not sold, used for advertising or used to train AI models without explicit permission.
Transparent and reversible
The individual can see what is held, correct it, export it, delete it and withdraw access at any time.
Respect for other people
People who appear in someone else's conversations or recordings have privacy rights too. Information about them should be minimised, not collected or analysed by default.
These are design objectives guiding the research and prototypes. They have not been independently verified.
Private, secure personal AI
Today's AI assistants retain little about the people who use them, and what they do retain usually sits with the provider. This research explores the opposite model: a personal intelligence foundation owned and controlled by the individual, which could organise memories, knowledge, experiences and authorised personal context across devices and AI systems without handing that information to a platform.
What is being built and tested
- A private, encrypted back-end store for personal memory and context, under the individual's control.
- Secure, authorised capture of conversations, notes and context from phones and wearable devices.
- Retrieval and organisation of personal memory over time.
- Connecting to different AI models without handing personal data to any single provider.
Open questions
- Consent, including for other people captured incidentally.
- Retention and deletion.
- Provenance of stored information.
- Which AI models may access what, and when.
- Portability between providers and devices.
- Where the boundaries of personal AI representation should sit.
Secure digital identity and AI representation
This strand asks whether selected personal knowledge and communication patterns could support an interactive digital representation of a person, for example one that answers questions in their style, with their authorisation.
It raises difficult questions. How can authenticity be established? Who may authorise, limit or revoke a representation? How is impersonation prevented, and how should a representation always identify itself as one? From a security perspective, a convincing representation of a person is also an attractive tool for fraud, so protection against misuse matters as much as capability. The research treats these as open questions rather than solved problems.
No claim is made that AI can capture or reproduce a complete personality, consciousness or identity.
Privacy-first wearable AI
Smart glasses with heads-up displays, discreet recorders and other wearables could become a natural, hands-free interface for personal intelligence. They also raise questions that phones largely did not, because they work at eye level, in shared spaces, and can be far less obvious to the people nearby.
Current work involves hands-on testing of wearable recording devices and smart glasses with heads-up displays, focusing on secure capture, processing on the device, and protecting the people nearby.
Protecting the privacy of others
The people around someone wearing these devices have not agreed to be recorded or analysed, so a responsible design has to protect them as well as the wearer. The research is exploring approaches such as:
- A clear, visible signal whenever capture is active.
- Capture that is off by default and started deliberately, rather than always on.
- Processing on the device, so raw audio and video are not sent elsewhere.
- Automatically minimising or discarding information about bystanders that is not needed.
- No facial recognition or identification of members of the public.
- Short retention and simple deletion.
Collaboration
I welcome conversations with manufacturers, developers and research partners working on smart glasses with heads-up displays, recording devices and other hands-free AI interfaces, particularly those who share a privacy-first approach. Open APIs, SDKs, on-device processing and strong security and privacy controls are priorities.
Evaluation devices
Manufacturers and developers are welcome to provide production or pre-release AI wearables for independent evaluation. Devices are tested hands-on, customised where the platform allows, and assessed for security, privacy, data handling and protection of bystanders, as well as day-to-day usefulness.
Findings can be shared privately with the manufacturer, and confidentiality agreements for pre-release hardware are respected. Evaluations remain independent: providing a device does not secure a favourable view.
No commercial partnership or endorsement is implied by any collaboration or evaluation.
Common questions
Is any of this research available as a product?
No. These are active research projects with working prototypes, built and tested privately. They are not products for sale.
Where is personal information stored?
In private, encrypted storage under the individual's control, not pooled on a provider's platform. Processing stays on the device or self-hosted systems wherever practical, and external AI services receive only the minimum needed, with authorisation.
Is personal information used to train AI models?
No. Personal information is not sold, used for advertising or used to train AI models without explicit permission.
How are people around a wearer protected?
Through approaches such as a visible signal when capture is active, capture that is off by default, processing on the device, minimising information about bystanders, and no facial recognition of members of the public.
Can manufacturers provide devices for evaluation?
Yes. Production or pre-release wearables are welcome for independent, hands-on evaluation. Please describe the device using the contact form first. Confidentiality agreements are respected, and evaluations remain independent.
Is this research part of PENWORTHIA?
No. This research is independent and is not connected to PENWORTHIA, which is a separate assurance platform for organisations.
Working on privacy-first personal AI, secure personal data storage or wearable technology? Collaboration enquiries are welcome.
Discuss research collaboration