Continuous care depends on trust. Micromeet Care Workspace connects medical check-up reports, care plans, follow-up and appointments. Cowork, the AI coworker in Care Workspace, prepares the work; clinicians and care teams decide the next step.
Your patient and institution data are handled on your instructions, with clear permissions and agreed data-processing terms.
Explore our data commitments, security controls, independently certified management systems and regional regulatory evidence.
Patient and institution data belong to the customer. Micromeet processes data only to deliver the contracted service, on customer instruction, and under the applicable data processing agreement.
You own your data
Patient, clinician, and institutional data remain customer data. Selling data is not part of Micromeet’s business.
No model training on identifiable data
Micromeet does not use identifiable patient or customer data to train AI models. Product improvement uses de-identified data only where the required consent and agreements are in place.
Third-party model restrictions
External model providers are used under enterprise terms designed to prohibit training on customer data, with retention minimized to what is needed to return a result.
Regional residency
Micromeet uses Alibaba Cloud for hosting and storage in Singapore, Indonesia and Hong Kong, and for AI processing in Singapore and Hong Kong. BytePlus provides AI processing in Malaysia. Each institution’s data is handled in the regions agreed for its deployment.
Retention and deletion
Retention is governed by customer agreements, service requirements, and legal obligations. Customer data can be deleted on request and at contract end according to the applicable agreement.
Clinical governance
AI writes. Doctors decide. AI-assisted clinical content is governed as clinician-in-the-loop clinical decision support.
AI writes. Doctors decide.
AI-generated clinical content must be reviewed and may be edited by the responsible physician before it is released, communicated, or relied on in a patient-facing workflow.
Clinical accountability by item or section
For composite workflows such as MCU reporting, responsibility can be assigned at the clinical item or report-section level, so each responsible physician remains accountable for their part of the output.
Audit trail and data provenance
The raw AI output, physician edits, reviewer identity, timestamps, release action, operator, and version history are retained as a traceable audit trail for quality review and incident investigation.
Clinical risk escalation
If AI-assisted content raises a material clinical concern, escalation follows the responsible physician or designated clinical escalation pathway rather than an autonomous AI decision path.
White paper summary
How we contain risk
Micromeet Care Workspace combines controlled cloud services, scoped access to customer systems and human approval for consequential actions. These controls are designed to limit what AI-assisted workflows can access and change.
HIPAA-aligned cloud security
Micromeet uses Alibaba Cloud infrastructure. Alibaba Cloud documents HIPAA compliance support and Business Associate Agreements (BAAs) for customers with HIPAA requirements. The services, deployment configuration and contractual arrangements are confirmed for each institution.
Cloud services run on managed infrastructure with encrypted transport and storage, monitored operations, controlled releases, and backup and recovery paths.
Controlled workflow access
Care Workspace workflows operate within customer-authorized permissions and approved integrations. AI assistance does not grant additional access to patient records or customer systems.
Protected data handling
Task data is processed only for authorized workflows. Where a deployment uses a local client, device-held credentials and cached data are protected through operating-system security controls and encrypted storage.
Containment principle
Skill boundaries, domain allowlists, and human confirmation contain what an influenced model can do.
Security layers
Layer 1
Cloud
Micromeet cloud services provide identity, policy, model orchestration, configuration release, logs, monitoring, storage, and operational support through controlled infrastructure.
WAF and edge filtering
Internet-facing services are protected with web application firewall controls, TLS, traffic filtering, and managed rules to reduce common web attack paths.
Scalable and reliable platform
Cloud capacity planning covers compute, storage, network, and model workloads so customer growth and peak task volume can be handled predictably.
Availability and recovery
Production services use health checks, multiple service instances, recovery procedures, and backup paths to reduce service interruption risk.
Role-based cloud operations
Administrative access is limited to authorized personnel, scoped by role, protected by least-privilege permissions, and logged for review.
Encrypted storage and backups
Databases, object storage, logs, and backup environments use cloud encryption capabilities, with access controlled by environment and role boundaries.
Monitoring and response
Cloud monitoring, application logs, access logs, security alerts, and failed-task signals support troubleshooting, audit review, and incident response.
Controlled release workflow
Build, deploy, and publish workflows produce evidence that supports change review, rollback, and operational accountability.
Layer 2
Care Workspace
Care Workspace uses controlled workflows, approved integrations and human review to limit the impact of model mistakes and prompt-injection attempts. Local-client controls apply where that deployment includes a client application.
Signed skill packs
For local-client deployments, capabilities are delivered through centrally authored, reviewed and signed skill packs. AI cannot grant itself new permissions at runtime.
Allowlisted network access
Local-client network access is restricted to approved customer systems, APIs and model endpoints. Requests outside the allowlist are blocked by policy.
Sandboxed client runtime
Where a local client is deployed, customer-system interactions are confined to authorized views and workflow actions. The client does not grant AI general control over other local applications.
Encrypted local storage
Where a local client is deployed, its cache and app data use encrypted on-device storage, with keys protected by operating-system security controls.
Credential protection
For local-client integrations, customer-system credentials stay in operating-system protected storage on the customer device, not in Micromeet backend services.
Human-approved writes
The agent proposes; a person confirms. Write or irreversible actions require explicit approval and are retained in the audit trail.
Replayable audit trail
Agent steps, key inputs and outputs, confirmations, operators, timestamps, and version changes are recorded for troubleshooting and compliance review.
Prompt-injection containment
Security does not depend on the model never being influenced. Skill boundaries, domain allowlists, and human confirmation contain what an influenced model can do.
Layer 3
Employee operations
Micromeet employees operate under confidentiality obligations, role assignment, documented procedures, and security lifecycle controls for regulated healthcare workflows.
Confidentiality handbook
Employees are required to follow the employee handbook, confidentiality expectations, and security handling requirements for customer, clinical, and operational data.
Role-based training
Security awareness, privacy handling, and workflow-specific training support consistent behavior in regulated healthcare environments.
Regulated workflow discipline
Operational changes, support handling, and escalation paths are performed through documented workflows with review, approval, and audit evidence.
Secure engineering lifecycle
Code changes, infrastructure configuration, and releases are reviewed through engineering workflows with automated checks and rollback paths where applicable.
Vulnerability management
Dependencies, images, cloud configuration, and public endpoints are reviewed through scanning and prioritized remediation.
Vendor management
Subprocessors are tracked by purpose, region, data category, security obligation, contract status where applicable, and vendor-review evidence, with customer-facing disclosure through the Trust Center or contract materials.
Subprocessor management
We disclose every third party that processes data on our behalf. Model providers process data under Micromeet's data-processing terms, which restrict use of customer data for training; hosting and processing locations follow each institution's agreed deployment. The authoritative subprocessor register is available on request.
Subprocessor
Purpose
Data processed
Region
Data terms
Alibaba Cloud — hosting
Cloud hosting & storage
Customer and patient data
Singapore, Indonesia and Hong Kong
Cloud hosting controls aligned to HIPAA security expectations.
Alibaba Cloud — AI services
AI processing
Task content sent for processing
Singapore and Hong Kong
Processed under Micromeet’s data-processing terms; not used to train Micromeet’s models. Provider terms and retention: authoritative register, available on request.
BytePlus — AI services (ModelArk)
AI processing
Task content sent for processing
Malaysia
Processed under BytePlus ModelArk terms; not used to train models. Authoritative register available on request.
OpenAI
AI processing, where used
Task content sent for processing
Singapore
Where used: under OpenAI API/enterprise terms that restrict training on customer data. Authoritative register available on request.
Illustrative list — the authoritative subprocessor register is maintained and available on request.
Regional regulatory programmes
Review Micromeet’s participation and results in local healthcare regulatory programmes.
Indonesia · Ministry of Health
Passed (Lolos)
Micromeet Passes Indonesia’s Regulatory Sandbox
Following testing, evaluation and guidance, Micromeet passed the 2025–2026 programme with a recommendation for Tanda Daftar Inovasi Teknologi Kesehatan (ITK).
Programme scheme
Skema A — Regulatory Sandbox
Official announcement
11 September 2026
Assessed feature, per final report
AI Medical Scribe feature supporting the Medical Check-Up (MCU) process.
Validity stated in the final report
September 2026 to March 2027.
The programme result applies to the feature and period stated in Micromeet’s final report. It does not certify the whole Care Workspace or Care Loop, and the ITK recommendation is not a market authorisation or an information-security certification.
The public Ministry announcement confirms the programme result. The assessed feature and validity above come from Micromeet’s programme-specific final report. Contact us for the applicable evidence during procurement or regulatory review.
Request trust documents for security review, procurement, and compliance diligence.
Frequently asked questions
Which AI models does Cowork use?
Cowork uses models including DeepSeek, Qwen and OpenAI models, through the AI services listed under Subprocessors: BytePlus (Malaysia), Alibaba Cloud (Singapore and Hong Kong), and OpenAI where used (Singapore). Each deployment uses the models and processing regions selected for it. External model providers are used under terms that restrict training on customer data. The authoritative subprocessor register is available on request.
Are our inputs and outputs used to train AI models?
No. Micromeet does not use identifiable patient or customer data to train AI models, and external model providers are used under terms that restrict training on customer data. Retention by providers is limited to what is needed to return a result.
What can a preset skill access or change?
A skill is a reviewed set of instructions for a task, such as preparing a check-up summary or a follow-up worklist. It works only with the data and apps your team has authorized for that task. It cannot give itself new permissions, and any write or irreversible action waits for a person to confirm it.
Who signs off clinical content prepared by Cowork?
The responsible physician. AI-prepared clinical content is reviewed, and may be edited, before it is released or communicated to a patient. Edits, reviewer, time and release are kept in the audit trail.
Can AI expand its own access?
No. Care Workspace operates within customer-authorized permissions and approved workflows. AI assistance does not grant itself additional access to records, integrations or actions.
Where does our data go?
Data processing follows the contracted deployment region and approved service providers. See our data commitments and subprocessor information for the applicable hosting and AI-processing arrangements. Task access is limited to the data needed for the authorized workflow.
Is our data encrypted?
Yes. Data is encrypted in transit with TLS 1.3 where supported and at rest. Where a local client is deployed, its on-device data uses encrypted storage with keys protected by operating-system security controls.
How is access to Care Workspace data controlled?
Access follows authorized user permissions and approved integration interfaces. Using Care Workspace does not grant other applications access to patient records. Customers also manage device security, user accounts and physical access to their systems.
Do you store our login credentials?
For local-client integrations, customer-system credentials stay on-device in operating-system protected storage and are not sent to Micromeet’s backend or model providers. Credential handling for other integrations follows the agreed deployment configuration.
What about prompt-injection attacks?
Workflow permissions, approved integration scopes and human approval are designed to limit the impact of an influenced model. Local-client deployments also use signed skill packs and network allowlists to constrain available actions and endpoints.
Who can perform write actions in our systems?
A human. Every write or irreversible action requires explicit user confirmation before the agent executes it.
How do you protect cloud services?
Cloud services use TLS, encrypted storage, role-scoped access, monitoring, WAF and edge controls, environment separation, backups, and controlled release workflows.
How are backups and recovery handled?
Production data is protected through encrypted backup environments and recovery procedures. Backup frequency, retention, RPO, and RTO can be provided in customer review materials where applicable.
What is the customer responsible for on local devices?
Customers remain responsible for device ownership, operating-system patching, disk encryption, endpoint protection, account lifecycle and physical access control. Where a local client is deployed, Micromeet provides its application controls, local encryption, signed updates, network allowlists and audit records.
How are vendors and subprocessors governed?
Subprocessors are tracked by purpose, region, data category, access need, and security obligation. The public list is maintained in this Trust Center and can be supplemented through contract materials.
Contact us
Contact support
Request trust documents, submit a security or compliance question, or report a vulnerability. Our team will route the request to the right owner.
