Vendor due diligence

Answers current as of 18 October 2026.

Every answer here is Aicura's current position, including where the answer is that Aicura doesn't do something yet. The answers describe the hosted Aicura service, the platform at aicura.app that Aicura runs for its customers, and they are drawn from our Terms of Service, Data Processing Agreement and trust pages. Where they differ, the Terms and the DPA prevail.

If you buy Aicura through a cloud marketplace, it runs entirely in your own cloud account instead, and no customer data, telemetry or support access reaches Aicura. The hosting, sub-processor and data handling answers below don't apply to that installation.

Questions about this page go to [email protected], privacy questions to [email protected] and security questions to [email protected]. If your organisation has its own questionnaire, send it to us and we'll fill it in.

1. Company and governance

What is the vendor's legal name and where is it incorporated?

Aicura Pty Ltd. (ABN 85 696 789 876), an Australian proprietary limited company based in Melbourne, Victoria, and registered on 1 April 2026. Aicura serves customers in every country from Australia.

Who runs the company?

Aicura is designed, built and run by its founder, Scott Goldie, who is responsible for the platform, its security and its handling of personal information.

Who is the contact for privacy and security matters?

Privacy enquiries go to [email protected]. Security reports go to [email protected], as published in our security.txt and on the security policy page. Legal notices go to [email protected].

Which privacy laws does the vendor operate under?

Aicura is an APP entity under the Privacy Act 1988 (Cth) and follows the Australian Privacy Principles. Where the EU GDPR or the UK GDPR applies to our handling of personal information, our Privacy Policy gives the information those laws require, and the DPA covers the personal information we process for customers.

2. Data location and hosting

Where is customer data stored?

Aicura has one hosting region today, in Australia, and every customer is served from it wherever the customer is based. Application servers, database, object storage, workflow engine and embeddings sit in AWS ap-southeast-2 (Sydney). Backups are written to AWS ap-southeast-4 (Melbourne), in a separate AWS account locked down for backup-only use. Customer document content, the AI register, generated outputs, support tickets, audit logs and AI processing all stay in Australia.

Where does AI processing happen?

Model calls run on Amazon Bedrock's Australian cross-region inference, so each one runs in Sydney or Melbourne. Embeddings are generated in Sydney.

Does any data leave Australia?

A few narrow flows do, and none of them carries the documents you upload or anything you haven't chosen to publish.

• Stripe processes billing data in the United States.
• Resend sends transactional and briefing emails from AWS Tokyo (ap-northeast-1).
• Cloudflare handles DNS, content delivery and the web application firewall on its global edge network, so inbound requests pass through it.
• Cloudflare R2 stores the Trust Centre pages you choose to publish, in its Asia-Pacific region.
• Cloudflare D1 stores submissions to the website's contact and briefings forms on its global edge.

What safeguards apply to transfers?

Section 11 of the DPA sets them out. Aicura takes reasonable steps under APP 8 to make sure overseas recipients handle personal information consistently with the Australian Privacy Principles. Where the GDPR applies to a transfer to Aicura, Module Two of the EU Standard Contractual Clauses is incorporated into the DPA, and where the UK GDPR applies, the UK Addendum is. Onward transfers to sub-processors rely on an adequacy decision or on the sub-processor's own transfer terms.

What stops data moving to another jurisdiction without notice?

The Terms of Service state that your data is hosted in Australia. Aicura gives at least 30 days' notice before adding or replacing a sub-processor. If you object and we can't resolve it, you may terminate and receive a pro-rated refund for the unused portion of your current term.

3. Use of customer data for model training

Is customer data used to train or improve AI models?

No. Your data is never used to train AI models, Aicura's or anyone else's. This is a commitment in the Terms of Service. Aicura reaches its models through Amazon Bedrock under terms that exclude your inputs and outputs from training of foundation models, and the model providers, including Anthropic, cannot access your prompts or completions.

Is there an opt-out?

There is nothing to opt out of, because customer content is never used for training.

Are copies of prompts kept anywhere else?

Aicura keeps a log of each AI call, including the prompt and the output, for up to 90 days for debugging and quality assurance, then deletes it automatically. AWS-side full-text invocation logging is deliberately not enabled, so no second copy of your content sits in AWS's logs.

4. Encryption and access controls

Is data encrypted in transit and at rest?

Yes. All traffic over public networks uses TLS 1.2 at minimum with TLS 1.3 preferred, and older versions are rejected at the load balancer. Inside the platform, connections to the database, to AWS services and to the workflow engine use TLS, and traffic between the load balancers and application servers stays on private, access-controlled networks. The database, object storage, secrets store and backup vault are encrypted with AES-256 using customer-managed AWS KMS keys with automatic annual rotation.

Who controls the encryption keys?

Aicura does. Keys are scoped per environment and per data class and are held in a dedicated AWS security account, separate from the application accounts. The Evidence Vault, the most sensitive store, has its own key. Customer-held keys are not offered on the hosted service.

How is each customer's data kept separate?

In two independent layers. In the database, PostgreSQL row-level security is enforced on every tenant-scoped table, and the role the tenant API uses cannot bypass it. In object storage, each request uses credentials pinned to a single tenant's storage area, so a wrong tenant identifier is denied by AWS IAM rather than by application logic.

How do users sign in, and is MFA required?

Authentication is handled by Auth0 in an Australian tenant, and passwords never touch Aicura's servers. Multi-factor authentication is enforced for administrative roles and prompted adaptively, based on risk signals, for everyone else.

Are access controls role-based?

Yes. A role grants read or write access to each area of the product, within what your plan includes, and can be narrowed to business units. Roles can be held through an identity provider group of the same name or assigned by an administrator. Every tenant starts with a built-in Administrator role that can't be edited or disabled, and always keeps at least one administrator. Every role change is captured in the Evidence Vault.

How do Aicura's own staff access production?

Through AWS IAM Identity Center with federated single sign-on and multi-factor authentication. There is no standing administrative access, no long-lived database credentials, no public database endpoint and no break-glass path to the production database. Changes operators make to your data are recorded in the audit log alongside changes your own users make.

Are access and change logs kept, and can we see them?

Yes. Every change to your organisation's records is written to a change log, recording who changed what and when, in the same database transaction as the change itself, and retained for 7 years. Your administrators can browse it in the Audit area of the product. Authentication events are exported to a dedicated storage bucket under S3 Object Lock in compliance mode for 7 years, where no one, including Aicura, can delete or modify them. AWS CloudTrail records all account-level activity into a separate audit account with log file validation enabled.

5. Certifications and independent assurance

Does the vendor hold SOC 2, ISO/IEC 27001 or ISO/IEC 42001?

No. Aicura holds none of these and has not been IRAP assessed, and we don't claim otherwise. Aicura is evaluating ISO/IEC 42001 as the framework for governing its own AI.

What assurance covers the infrastructure?

Aicura runs on AWS, whose attestations and certifications for the services and regions Aicura uses are available to you through AWS Artifact under AWS's terms.

Has the vendor had an independent penetration test?

Not yet. Aicura has not commissioned an independent penetration test of the platform.

How are vulnerabilities and threats detected?

The platform runs under AWS GuardDuty for threat detection, Inspector for vulnerability scanning of compute and container images, Security Hub for control monitoring, Config for configuration drift and CloudTrail for account activity. Infrastructure reviews produce a tracked list of improvements that we work through. Anyone can report a vulnerability to [email protected].

6. Incident response and breach notification

How quickly will we be told about a breach?

Without undue delay and in any event within 24 hours of Aicura becoming aware of an actual or reasonably suspected breach affecting your personal information. A full incident report follows within 72 hours of the breach being confirmed. These commitments are in section 10 of the DPA.

What does the notification contain?

As far as it is known at the time, the nature of the breach including the categories and approximate number of people and records affected, its likely consequences, the measures taken or proposed to address it, and a named person at Aicura for further information.

Who notifies regulators?

Where Aicura itself has a notification obligation, under the Notifiable Data Breaches scheme or another law that applies to us, Aicura notifies the regulator directly within the time that law requires. Your organisation remains responsible for its own notifications, and Aicura cooperates with your investigation and with those notifications, including under Articles 33 and 34 of the GDPR and UK GDPR.

Where are service incidents published?

On the status page, which Aicura updates when an incident or scheduled maintenance affects the platform, with a record of past incidents.

7. AI models and transparency

Which models does the platform use?

Anthropic's Claude models, through Amazon Bedrock in Australia. As of 18 October 2026, Claude Haiku 4.5 handles light tasks, Claude Sonnet 4.6 handles standard tasks and Claude Opus 4.6 handles the heaviest. Embeddings use Amazon Titan Text Embeddings V2 in Sydney. The model used for every call is recorded in the AI call log.

What are the models used for?

Reviewing the privacy and governance documents you upload, drafting disclosures and policy text from your register data, pre-populating risk assessment fields and reading the helper documents you upload. The models are not used to make compliance determinations, give legal advice or produce final outputs without human review.

Where is the model documentation?

Anthropic publishes model cards for each Claude model, covering training, capabilities, limitations and safety evaluations, and they are the authoritative source at the model level. Aicura does not train or modify the models. Aicura's interpretive content, meaning its assessment briefings, framework mappings and regulatory criteria, is curated by Aicura, held in the platform and applied through prompts.

Are the models used to make decisions about individuals?

No. Aicura uses them for document analysis and structured drafting, not for decisions about people, code execution, autonomous action or tool use.

8. Accuracy and human review

How does the vendor manage inaccurate or fabricated output?

Every AI-generated output is a draft that a person must review before anyone relies on it, shares it or publishes it. AI-generated content is labelled in the product, and the human review requirement is shown at the point of generation. Workflows that affect your published documents and governance record include review steps where your people accept, edit or send the output back.

What does the vendor say the platform is not suitable for?

Aicura is a support tool for AI governance work. It does not give legal advice, certify or audit anything, or guarantee compliance with any law or framework, and its outputs are not a substitute for your own counsel and decision-makers. The boundaries page sets out what Aicura does and doesn't do.

Does the platform publish anything itself?

No. Aicura drafts. Your policy goes live through your own publishing process, and a Trust Centre page or guardrail release goes live only when one of your people publishes or deploys it.

9. Model changes and monitoring

How are model changes handled?

Aicura calls Bedrock through its Converse API, which uses one request format across model families, so moving to a different Bedrock-hosted model, or back to an earlier one Bedrock still offers, is a configuration change followed by a test, not a rebuild. Aicura's prompts already run against non-Claude models in development.

What happens if a model concern is disclosed?

Aicura assesses the impact on its specific use, notifies affected customers and switches to an alternative Bedrock-hosted model if the concern materially affects that use.

What is monitored for each AI call?

The AI call log records the function, provider, model, prompt, response, token counts and latency of every call. Aicura does not train its own models, so drift in the sense of a retrained model degrading does not arise.

Does the model choice concentrate our risk on one AI vendor?

Not materially. The register, assessments, generated documents, workflow and interpretive content are Aicura's, and the model underneath can be substituted. On your own register, the right entry is the Aicura platform running on Amazon Bedrock in Australia, rather than the underlying model.

10. AI safety controls

What stops harmful output or prompt injection?

The prompts that drive every AI call are written by Aicura, not by your users, and there is no free-form prompt surface. Every input is structured data Aicura builds or a document your organisation chose to upload, and document content is treated as data, not as instructions. The model has no authority to take actions, and its output goes to human review.

Is a model guardrail applied?

A model guardrail capability exists but is deliberately not applied as a generic filter. Generic content and topic filters misfire on regulatory text, and PII redaction would strip the contact details that generated privacy and governance policies are meant to contain. The responsible AI page explains the reasoning in full.

What about the AI gateway and scanners we run?

They run inside your own environment, and your AI traffic never reaches Aicura. The Hub you run connects outbound to Aicura, so you open no inbound firewall rule and issue Aicura no credentials into your systems.

Can generated content be removed?

Yes. Your users can delete generated content in the product. AI call logs are deleted automatically after no more than 90 days.

11. Audit rights

Can we audit the vendor?

Aicura makes available the information reasonably necessary to show it meets the DPA, including this page, the DPA, the Privacy Policy, the Terms of Service, the trust pages, AWS's attestations through AWS Artifact and written answers to your specific questions. Direct on-site or remote audits are not offered as standard. Where a privacy law, the Standard Contractual Clauses or a regulator with jurisdiction over you requires more, we agree the scope, timing and cost with you in good faith. Section 12 of the DPA sets this out.

12. Warranties and liability

What does the vendor warrant?

That the service is provided with the due care and skill expected of a competent provider of similar services. Aicura does not warrant that the service will be uninterrupted or error-free, that its outputs will result in compliance with any law or framework, or that AI-generated content will be accurate, complete or fit for your circumstances.

What is the liability cap?

To the extent the law allows, Aicura's total liability is capped at the fees you paid in the 12 months before the event giving rise to the claim.

What is excluded?

To the extent the law allows, indirect and consequential loss, loss of profits, revenue, business or opportunity, loss of goodwill, loss of or corruption to data, regulatory fines and penalties, and legal costs or third-party claims arising from reliance on the service. Liability that can't lawfully be excluded, including for death or personal injury caused by negligence and for fraud, is not excluded, and nothing removes consumer rights that can't be excluded under the law that applies to you.

Who is responsible for decisions made using AI output?

Your organisation is. All AI-generated content is a starting point that you review, edit and approve before relying on it. Clauses 3, 10 and 11 of the Terms of Service set this out.

13. Availability and support

Is there an uptime commitment?

The standard Terms of Service carry no uptime commitment and no service credits. Where an availability commitment exists, it sits in your Enterprise agreement. The status page records incidents and maintenance.

How do we get support?

Open a support ticket from inside the product, or email [email protected]. Aicura responds during Melbourne business hours.

14. Data portability and exit

Can we get all our data back in a standard format?

Yes. On request, Aicura provides an export as a ZIP archive containing JSON files for each table plus your original uploaded files, as soon as practicable and in any case within 30 days. The product also lets you access, correct, export and delete your data yourself. Standard assistance is included at no extra charge.

What happens to our data when we leave?

Aicura keeps your tenant data for up to 60 days after your subscription ends, so you can ask for an export or reactivate, then deletes it. You can ask for deletion sooner by emailing [email protected], and we aim to delete within 30 days. A limited set of records, meaning audit logs, billing records and authentication audit logs, is kept for up to 7 years where the law requires it, to defend legal claims or for tax purposes, and is not otherwise used.

What if the vendor stops offering the service?

Aicura can end the Terms on 30 days' written notice, and the 60-day export window above still applies after that.

15. Intellectual property

Who owns our data and the outputs?

You own the data you upload and the content you create with the service. You can use, edit and adapt the drafts Aicura produces for you in your own materials. The underlying templates, prompts and regulatory content remain Aicura's.

Who owns the platform and the models?

Aicura owns the platform. Aspects of it are the subject of pending Australian patent applications 2026905992 and 2026906023. The Claude models belong to Anthropic and the Titan embeddings model to Amazon, and Aicura uses them through Amazon Bedrock.

Does the vendor give an IP indemnity?

The standard Terms of Service do not include an intellectual property indemnity from Aicura.

What happens if the vendor is acquired?

Aicura can assign the Terms to a successor in a sale, merger or restructure of its business, so your agreement and your data handling commitments carry over.

16. Sub-processors and supply chain

Which third parties process customer data?

The sub-processors page is the authoritative, current list, with what each one sees. GitHub Actions builds Aicura's software and processes code only, so no customer data reaches it. There are no advertising, analytics or data enrichment services.

Will we be told about changes?

Yes, at least 30 days before a new or replacement sub-processor takes effect. If you object on data protection grounds and the objection can't be resolved, you may terminate and receive a pro-rated refund for the unused portion of your current term.

Do sub-processors carry the same obligations?

Aicura imposes on each sub-processor written obligations equivalent in substance to those in the DPA, and remains responsible for their acts and omissions as if they were its own.

For your own assessment

These questions are about your organisation's readiness to adopt any AI system, not about Aicura. They are left for your team to answer, and the table below records your view of each section of this page.

1. Does your organisation have an AI governance framework, executive sponsorship and documented accountability for AI systems?

2. Do you have the people (security, privacy, risk, technical) to manage the system and the vendor relationship?

3. Have you assessed the potential harms of the system for your use and the mitigations it needs?

4. Is there a process for monitoring the system and escalating issues with it?

5. What training do staff need to use the system responsibly and understand its limitations?

6. Will you start with a pilot, and what must be true before you expand from it?

7. What is your plan if the system does not meet expectations?