AI Usage Policy

Version 2.0 — May 2026

This Policy supersedes the AI Usage Policy v1.0 published in May 2026. The substantive principles remain unchanged; v2.0 expands the Policy to address automated decision-making rights, training data and model provenance, bias and fairness commitments, third-party processor relationships, cross-border data processing implications, governance and accountability, and AI incident response.

1. Why this Policy exists

Artificial intelligence is a meaningful part of the Circlworld platform. We use AI to help treasurers manage circles, support members with everyday questions, generate analytical insights, draft standardised documents, and improve operational efficiency. We use AI internally for triage, anomaly detection, and content generation under engineering supervision.

We believe AI used carefully makes financial coordination more accessible — particularly for people who would benefit from professional-level guidance but cannot afford individual professional advice. We also believe AI used carelessly can mislead, exclude, or harm — and that this risk is particularly acute for the communities Circlworld is built to serve, who have historically been underserved by formal systems and are most vulnerable to systems that get things wrong.

This Policy explains the principles, the limits, the legal framework, and the protections we apply. It is in addition to our Terms of Service, Privacy Notice, Data Protection Policy, Consumer Protection Policy, and (for treasurer partners) the Treasurer Partner Agreement.

2. Five governing principles

Our AI usage is governed by five principles. Every concrete design choice elsewhere in this Policy traces back to one of these.

2.1 Information, not advice

AI on Circlworld provides information and analysis, not financial, legal, or other regulated advice. The distinction is structural, not cosmetic:

AI describes options, trade-offs, considerations, and data
AI does not recommend specific actions, decisions, or products
AI does not replace professional advice from qualified human advisers
All AI outputs carry disclaimers clarifying their informational nature

This boundary protects you. AI that gives confident-sounding advice about financial decisions can be wrong, mislead you, and cause genuine harm. AI that provides structured information helps you decide for yourself.

2.2 Human decisions, not AI decisions

AI on Circlworld supports decisions; it does not make them. Decision-makers are:

Members, when joining circles, upgrading to Pro, sharing Trust Reports, or taking any action affecting their own interests
Circle members collectively, when deciding circle terms, accepting new members, voting on LLP or CLP requests, or resolving disputes
Treasurers, when administering circles, confirming contributions, managing disbursements
Circlworld staff, when reviewing KYC, handling complaints, or taking platform-level actions

AI surfaces relevant information for these decisions. It does not make them.

2.3 Transparency about AI

When you interact with AI on Circlworld, we tell you:

That AI is involved
The tier of AI (Basic, Enhanced, or Enterprise)
The general nature of the analysis being performed
The limitations of the output

You are never in a situation where you interact with AI without knowing.

2.4 Human oversight

AI usage on the platform is overseen by humans:

Engineers monitor AI behaviour for unexpected or problematic outputs
Compliance staff review AI behaviour quarterly in regulated contexts
Member-facing AI behaviour is reviewable through our complaints process
A designated AI Governance Lead is accountable for AI behaviour across the platform (see Section 13)

We do not deploy AI in any context where decisions affecting your legal rights are made without human review.

2.5 Constrained capability

AI on Circlworld is constrained to specific tasks. We do not deploy general-purpose AI that can engage in any conversation. We deploy task-specific AI configured for particular operational support roles, with system-level constraints that prevent it from acting outside its designated scope.

3. Where we use AI

3.1 AI Assist — three tiers

Basic AI Assist (Free users £3/query; Pro users unlimited)

Treasurer operational guidance (bank account setup, communication templates)
Member onboarding Q&A
Contribution flow troubleshooting
Trust Report basic explanation
Dispute initial triage and routing

Enhanced AI Assist (Pro users £7/use or 1 credit; Business unlimited)

Circle health analysis with written assessment
Personalised treasurer coaching
Dispute resolution mediation suggestions
Member retention analysis
Cross-circle benchmarking
Trust report narrative generation
LLP borrower request analysis
CLP borrowing circle profile reading
Voting deliberation prompts

Enterprise AI Assist (Business users £35/use; Enterprise unlimited)

Portfolio analytics across multiple circles
Aggregate default risk descriptors (at circle level, never individual)
Regulatory reporting drafts
Cross-corridor compliance checking
API-accessible AI for institutional partners

3.2 Other AI applications

Standardised document drafting — AI drafts standardised LLP/CLP transaction documents (£15/document) including loan agreements, mutual aid agreements, and standard amendments. AI does not draft custom or bespoke legal documents.

Trust Report narrative generation — when generating Enhanced Trust Reports, AI may produce narrative summaries based exclusively on audit log data, reviewed by the member before sharing.

Internal operational AI — KYC review triage (flagging unusual cases for human review), audit log anomaly detection, internal documentation generation, code generation with engineer review, support ticket categorisation.

3.3 Where we deliberately do NOT use AI

We do not use AI for:

Credit decisions at the individual level (no scoring, no default prediction, no recommendations)
Financial advice (information only, never advice)
Sensitive personal characteristic inference (race, ethnicity, sexual orientation, political views, health, religious beliefs — prohibited by our Privacy Notice)
Final identity verification decisions (KYC providers' AI is reviewable by humans)
Automated content moderation (humans make final removal/suspension decisions)
Hiring or employment decisions
Insurance underwriting
Targeted advertising (we don't run targeted advertising)

4. The AI models and providers we use

4.1 Primary provider — Anthropic

Our AI capabilities are provided by Anthropic, the company that produces Claude. We use Anthropic because:

They publish a Constitutional AI framework aligned with our principles
They have a strong safety record relative to industry peers
They offer enterprise terms supporting the regulatory boundaries we require
They commit contractually that customer data is not used for model training

4.2 Models in use

Claude Haiku for Basic AI Assist (cost-effective for simple operational queries)
Claude Sonnet for Enhanced AI Assist (balanced cost and analytical quality)
Claude Opus for Enterprise AI Assist (highest quality for institutional analysis)

4.3 Training data and model provenance

We have not trained Claude. Claude was trained by Anthropic on a corpus of publicly available web data, licensed material, and other sources documented in Anthropic's published model documentation.

We acknowledge that:

Claude's training data may include copyrighted material; this is the subject of ongoing legal and policy debate
Claude's training data inevitably reflects the biases of its source corpus
Claude's training data has a knowledge cutoff date that affects what the model knows about recent events
Claude does not understand the specific context of your community, jurisdiction, or circle unless we explicitly provide that context in the prompt

For the most current information on Claude's training, capabilities, and limitations, please refer to Anthropic's published documentation at anthropic.com.

4.4 Future provider relationships

We may, in future, integrate AI from additional providers. Before doing so, we will:

Conduct due diligence on the provider's training practices, safety record, and contractual commitments
Update this Policy to reflect the new relationship
Notify members in advance via our usual channels
Maintain the same principles and boundaries described in Section 2

5. AI accuracy, limitations, and hallucination

5.1 What AI can and cannot do reliably

AI models, including Claude, are pattern-matching systems trained to produce plausible responses. They are not knowledge databases, they are not reasoning engines in the strict sense, and they do not have access to real-time information unless explicitly provided.

This has important practical implications:

AI can produce confident-sounding statements that are factually wrong. This is called "hallucination." It happens. We design the platform to reduce hallucination risk, but we cannot eliminate it.
AI works best within its trained scope. Edge cases and novel situations may be handled less well.
AI can perpetuate patterns from its training data, including biases and dated information.
AI does not know things that happened after its training cutoff unless we provide that information in the prompt.
AI does not know your specific situation the way a human relationship would.

5.2 How we mitigate AI errors

System prompts constrain output scope. AI is configured to refuse to answer questions outside its designated role.
Disclaimers are mandatory. Every AI output carries a disclaimer.
Critical AI outputs are reviewed by humans in sensitive contexts (LLP/CLP analysis, dispute mediation, regulatory reporting).
Audit logging records AI behaviour. Patterns of problematic output trigger investigation.
Users can flag AI errors through Settings → Support → Report AI Issue.
Quarterly behaviour reviews examine AI output samples in sensitive areas.

5.3 What to do if AI is wrong

If you believe an AI output is incorrect:

Do not rely on it for a decision. Use it as one input among many.
Verify against authoritative sources — your bank, lawyer, accountant, the platform's audit log, regulatory websites.
Report the issue at accessibility@circlworld.com or via Settings → Support → Report AI Issue.
For consequential errors (errors that caused you or others to make incorrect decisions), raise a complaint via complaints@circlworld.com. We will investigate and provide remedies where appropriate.

6. Bias and fairness

6.1 Why this matters

The communities Circlworld serves — Caribbean, sub-Saharan African, diaspora communities — are over-represented in datasets where AI systems have historically shown bias. Facial recognition systems mis-recognise darker-skinned faces at higher error rates. Language models can produce stereotyped output when discussing certain ethnic, national, or cultural contexts. Credit scoring AI has been shown to disadvantage communities Circlworld is built to serve.

If we deploy AI carelessly, we replicate the exclusion that motivates the platform's existence. We take this seriously.

6.2 Our commitments

No demographic inference. We do not use AI to infer race, ethnicity, sexual orientation, religion, political view, or other sensitive characteristics from your data or behaviour.
No demographic input into operational AI. Where AI surfaces analysis, the inputs are participation behaviour and platform events — not demographic characteristics.
Cultural context in prompts. Where AI processes culturally-specific information (pardna, susu, ajo, family obligations, religious holiday contributions), we configure prompts to recognise and respect these contexts rather than treating them as anomalies.
Bias monitoring. We sample AI outputs quarterly for evidence of differential treatment by region, language, or community type. Material patterns are investigated.
Member feedback channels. Members can report perceived bias in AI outputs. Reports are reviewed by the AI Governance Lead and may trigger system prompt updates.

6.3 Limitations of these commitments

We cannot eliminate bias entirely. Claude was trained on a corpus that reflects the biases of its sources. Our system prompts can shape behaviour but cannot fully overcome training data patterns. We commit to continuous improvement; we do not commit to perfection.

If you experience AI behaviour you believe to be biased, please report it. Your feedback materially improves the platform.

7. Automated decision-making rights

7.1 The legal framework

Under the UK Data Protection Act 2018 and UK GDPR (Article 22), the Jamaica Data Protection Act 2020, and equivalent frameworks in other jurisdictions, you have rights regarding automated decisions that:

Are based solely on automated processing (no meaningful human involvement)
Produce legal effects concerning you, or
Similarly significantly affect you

These rights include:

The right to obtain human intervention
The right to express your point of view
The right to contest the decision
The right to receive meaningful information about the logic involved

7.2 How Circlworld's AI usage interacts with these rights

The following Circlworld uses of AI may constitute automated decision-making within the meaning of Article 22:

| AI use | Solely automated? | Significantly affects you? | Article 22 applies? | |--------|-------------------|----------------------------|---------------------| | AI Assist suggestions | No — human acts on output | Variable | No | | KYC provider AI | Partially automated | Yes (account decision) | Yes — see safeguards | | Anomaly flagging in audit log | Partially automated | Indirectly | Triggers human review | | Activation Bonus qualification | Algorithmic, not AI | Yes (£5 bonus) | Algorithmic determination | | Identity Level assignment | Algorithmic, not AI | Yes (tier benefits) | Algorithmic determination | | Community Health Multiplier | Algorithmic, not AI | Yes (residual %) | Algorithmic determination |

7.3 Your rights and our safeguards

For any automated or AI-assisted decision affecting you, you have the right to:

Request human review. Contact privacy@circlworld.com or Settings → Privacy → Request Human Review.
Receive an explanation of the logic of the decision in non-technical language.
Contest the decision and submit your own evidence or perspective.
Receive a decision from a human if you do not accept the automated outcome.

We will respond to such requests within 20 business days, with extension to 40 business days for complex cases (with notice).

The safeguards we apply:

All KYC rejections include human review on request before becoming final
All Programme suspensions are reviewed by a human Compliance officer
All disputed Activation Bonus qualifications are reviewed by a human Programme officer
All Community Health Status determinations are reviewable on request

7.4 What we do not do

We do not make decisions solely on AI output that:

Terminate your Circlworld account
Suspend your participation in active circles
Confirm or deny a payout
Resolve a dispute against you
Issue a Trust Report determining your standing

These decisions, where they occur, always involve meaningful human review.

8. Data handling for AI processing

8.1 What gets sent to AI

When you interact with AI Assist:

Your query (text you typed)
Relevant context from the platform (your circle data, trust profile elements, transaction history) where necessary for the query
System prompts that configure the AI's behaviour
Conversation history within the session

We minimise context to what is necessary for the query. We do not send your full profile when only a small slice is needed.

8.2 What does NOT get sent to AI

Your password or authentication credentials
Your raw identity verification documents (passport scans, biometric data)
Other members' personal information beyond what is necessary to your query
Sensitive personal characteristics (race, ethnicity, sexual orientation, religion, political views) — we don't collect these
Payment card data (PCI scope is separated)

8.3 Cross-border AI processing

This is a material disclosure that affects your data protection rights:

Anthropic processes Claude API queries on infrastructure located primarily in the United States. This means that when you use AI Assist, your query and the platform context sent with it transit to and are processed in the United States.

Under UK GDPR and Jamaica DPA 2020, this is an international data transfer requiring an appropriate transfer mechanism. We rely on:

The UK International Data Transfer Agreement (IDTA) or UK Addendum to the EU Standard Contractual Clauses for transfers from the UK
Equivalent contractual protections for transfers from Jamaica
Anthropic's published commitments on data handling

If you do not consent to this international transfer, you can opt out of AI Assist features. Some platform features that depend on AI will become unavailable, but core platform functionality (joining circles, contributions, trust profile, Trust Reports without AI-generated narrative) remains available.

8.4 Data retention by Anthropic

Anthropic's published policy specifies retention periods for API queries. We do not retain your AI queries beyond what is necessary for:

The active session
Audit logging (queries logged in the platform audit log, retained per our Privacy Notice)
Quality investigation (samples reviewed by AI Governance for accuracy and bias)

We do not retain AI queries for training purposes. Anthropic does not use Circlworld member queries to train their models — this is contractually specified.

8.5 Third-party AI as data processor

Anthropic acts as a data processor under our control for the purpose of AI Assist functionality. The processor agreement specifies:

Purpose and scope of processing (limited to query response)
Security requirements (encryption in transit and at rest)
Subprocessing restrictions
Audit rights
Breach notification (Anthropic notifies Circlworld; Circlworld notifies you per our Privacy Notice)
Return or deletion of data on termination

Our Data Protection Policy lists all data processors including Anthropic.

Using AI Assist on Circlworld requires your consent, which you provide:

Implicitly, by using a free AI Assist feature (Pro Basic AI Assist)
Explicitly, by purchasing a per-query AI use (Basic for Free users, Enhanced for Pro users, Enterprise for Business users)
Continuously, by remaining on the platform after being informed of changes via Section 14

You can withdraw AI consent at any time:

For specific uses — decline the AI Assist option when offered (use the human alternative)
Account-wide — Settings → Privacy → AI Processing → Disable AI Assist

Disabling AI Assist:

Stops your queries being sent to Anthropic
Removes AI suggestions from your dashboards
Disables AI-generated Trust Report narratives (you can still generate plain Trust Reports)
Does not affect platform features that don't require AI

9.3 What if you've already used AI?

Past AI queries are retained in your audit log per our Privacy Notice. You can request deletion of AI query records (subject to audit integrity constraints — some records are retained because deleting them would prevent us from accurately reconstructing past circle activity).

10. Your rights regarding AI

10.1 Right to know

You have the right to know:

When you are interacting with AI
What data about you is being processed by AI
What the AI's output is based on (in non-technical terms)
The limitations and uncertainty of AI outputs

10.2 Right to human review

You have the right to:

Request human review of any AI-assisted decision affecting you
Speak with a human Circlworld staff member instead of AI for complaints, significant queries, or sensitive matters
Refuse AI processing of your data in any optional context

10.3 Right to challenge

You have the right to:

Challenge AI-generated content you believe is incorrect
Request correction of factual errors
Request removal of AI-generated content that misrepresents you (subject to audit log integrity)
Escalate AI-related concerns through our complaints process

10.4 Right to opt out

You can opt out of optional AI features. You cannot opt out of features that are integral to core platform operation (for example, KYC provider AI for identity verification), but you can choose not to use platform features that rely on AI Assist.

10.5 Right to explanation

For any automated decision affecting you significantly, you have the right to meaningful explanation in non-technical language. Request via privacy@circlworld.com.

10.6 Exercising these rights

To exercise any AI right, contact us:

Email: privacy@circlworld.com (for data protection aspects) or ai@circlworld.com (for AI behaviour aspects)
Platform: Settings → Privacy → AI Rights
Post: to the relevant Circlworld entity

We respond within 30 days (extendable to 90 days for complex requests, with notice).

11. AI in specific platform contexts

11.1 AI in disputes

When a dispute is filed in a circle:

AI may surface relevant context from the audit log to help mediators
AI does NOT make findings of fact
AI does NOT determine who is right
AI does NOT impose remedies
Human mediators or circle members make all substantive decisions

Both parties to a dispute have the right to know whether AI was used in the mediation process and what the AI surfaced.

11.2 AI for treasurer partners

Treasurer partners using AI Assist in administering circles should note:

AI suggestions are information for your decision-making, not directions
You are responsible for circle administration decisions, not the AI
AI-generated communications (drafted messages, dispute responses) should be reviewed before sending
AI does not replace the judgement and relationships that make community-building work
The platform's audit log records when AI was used and for what purpose, supporting transparency with circle members

11.3 AI in Trust Reports

When AI generates narrative summaries for Enhanced Trust Reports:

The narrative is based exclusively on factual audit log data
You review and approve the narrative before it is finalised
You can edit or reject the narrative
The narrative is marked as AI-assisted in the Trust Report

You can request a Trust Report without AI narrative at any time.

11.4 AI in lending coordination (LLP / CLP)

When AI is used for LLP borrower analysis or CLP profile reading:

AI summarises participation history and surfaces relevant patterns
AI does NOT recommend whether to lend
AI does NOT score the borrower's creditworthiness
AI does NOT predict default probability for individuals
All lending decisions are made by the circle's collective vote

Members of a lending circle can request the underlying audit log data behind any AI summary.

11.5 AI for institutional partners (Lender API)

Institutional partners using the Lender API to access Trust Reports may use AI in their own systems to process the verified data we provide. The data Circlworld provides is what it is — verified, cryptographically signed, sourced from the audit log. How the institutional partner processes that data is their responsibility under their own AI governance.

Circlworld does not control or take responsibility for partner AI systems. Partner agreements include:

A requirement that the partner discloses to its own users when AI is used
A prohibition on the partner using Circlworld Trust Report data to train AI models
A requirement that the partner's AI usage complies with applicable AI regulation in its jurisdiction

12. AI used by you

If you use Circlworld, you may also use AI on your end. We have no objection to this — AI is increasingly normal in how people work. However:

You remain responsible for what you submit to Circlworld, whether you used AI to generate it or not
AI-generated content that violates our policies (false attestations, sham circles, fraudulent disputes) is treated the same as any other violation
We may detect AI-generated patterns in user content; this is one routine signal we use alongside others, not a basis for automatic action
We do not prohibit AI-assisted communication with the platform or with other members
You should not represent AI-generated content as exclusively your own in contexts where authorship matters (treasurer endorsements, peer reviews)

13. AI governance and accountability

13.1 The AI Governance Lead

Circlworld has a designated AI Governance Lead, responsible for:

Overseeing AI behaviour across all platform features
Reviewing AI Usage Policy updates
Responding to material AI incidents
Approving new AI deployments before launch
Quarterly review of AI performance, accuracy, and bias metrics
Liaising with external auditors and regulators on AI matters

The current AI Governance Lead is the platform founder, Drew St'Clair. As Circlworld scales, this role may be allocated to a specialist hire.

Contact the AI Governance Lead: ai-governance@circlworld.com

13.2 Governance structure

AI Governance Lead — strategic accountability and material decisions
Engineering team — implementation, system prompt configuration, monitoring
Compliance team — regulatory review, complaint investigation
External counsel — periodic review of legal compliance
External auditors — annual review of AI behaviour (commencing Year 2)

13.3 Decision-making

Material AI decisions (new AI capability, change to system prompts in regulated areas, response to AI incidents) require:

AI Governance Lead approval
Engineering team implementation
Compliance team sign-off where regulatory implications exist
Member notification before deployment where the change materially affects users

13.4 Documentation

We maintain internal documentation of:

Current AI system prompts and configurations
Change history of AI behaviour
Incident log
Quarterly review reports
External audit reports

This documentation is available to regulators on request and forms part of our regulatory engagement framework.

14. AI incident response

14.1 What constitutes an AI incident

We treat the following as AI incidents requiring formal response:

AI producing factually incorrect output in a regulated context (financial, legal, medical)
AI producing biased or discriminatory output
AI behaving outside its designated scope
AI causing demonstrable harm to a user
AI being used by malicious parties to attack the platform
AI processor (Anthropic) suffering a security incident affecting our data

14.2 Response process

Detection — by user report, monitoring alert, or audit log review
Containment — immediate action to stop ongoing harm (disable specific AI feature, isolate affected accounts)
Investigation — by the AI Governance Lead with Engineering and Compliance support
Notification — to affected users within 72 hours where the incident affects their rights
Remediation — fix the system, compensate affected users where appropriate, update policies
Documentation — incident logged with root cause analysis
Disclosure — in our annual transparency report (when published) and to regulators where required

14.3 Reporting an AI incident

If you believe an AI incident has occurred:

Email: ai-incident@circlworld.com
Through the platform: Settings → Support → Report AI Incident
Through complaints: complaints@circlworld.com

We acknowledge within 24 hours and respond substantively within 7 days for incidents with potential user impact.

15. External audit and transparency

15.1 Internal review

We conduct internal AI review:

Continuous monitoring by engineering and compliance teams
Quarterly review by the AI Governance Lead with sampled output analysis
Annual review of overall AI behaviour against our principles

15.2 External audit

Beginning Year 2 of operation, we will commission annual external audit of:

AI behaviour against this Policy
Bias and fairness in AI outputs
Compliance with applicable AI regulation
Effectiveness of AI Governance framework

Audit reports will be summarised in our annual transparency report.

15.3 Red-teaming

We engage external red-teamers to test AI systems for:

Prompt injection vulnerabilities
Output manipulation
Bias surfacing
Boundary violations

Red-team findings inform Policy and system prompt updates.

15.4 Transparency reporting

We commit to publishing an annual AI Transparency Report (commencing Year 2) covering:

AI usage volume by tier
Material AI incidents and responses
Bias and fairness review findings
External audit summary
Policy updates
User reports of AI issues and resolution outcomes

16. Alignment with regulatory frameworks

16.1 Current alignment

This Policy is designed to be consistent with:

UK Data Protection Act 2018 and UK GDPR Article 22 (automated decision-making)
Jamaica Data Protection Act 2020
EU AI Act principles (insofar as they apply to our processing)
NIST AI Risk Management Framework (informal alignment)
OECD AI Principles

16.2 Emerging frameworks

We monitor and prepare for:

UK AI Bill (in development)
EU AI Act detailed implementation (2026-2027)
Jamaica AI regulatory framework (emerging)
US state AI laws as US expansion occurs

When new frameworks come into force, we will update this Policy and our practices accordingly.

16.3 AI Act risk classification

Under the EU AI Act risk framework:

Our use of AI for member-facing operational guidance is limited risk (transparency obligations)
Our use of AI for credit-related analysis is carefully designed to stay outside the high-risk category (we provide information for human decision-makers, not automated credit decisions)
We do not use AI for any prohibited purposes (social scoring, manipulation, real-time biometric identification in public spaces)

We do not categorise ourselves as a high-risk AI system provider under the EU AI Act. If the platform's AI usage evolves in a way that would trigger high-risk classification, we will comply with the corresponding obligations.

17. AI marketing and partner restrictions

17.1 How we represent AI capability

In our marketing and platform copy, we commit to:

Not overstating AI capability
Not creating false expectations about AI accuracy
Not implying AI is a substitute for professional advice
Not using AI imagery suggesting human-like understanding ("AI will care about you," "AI knows you")
Disclosing AI involvement where relevant

17.2 Partner restrictions on AI usage

Partners using our Lender API agree to:

Disclose to their own users when AI processes Circlworld data
Not use Circlworld Trust Report data to train their own AI models
Comply with AI regulation applicable in their jurisdiction
Not represent Circlworld AI outputs as their own analysis
Not use Circlworld AI outputs to make solely-automated credit or insurance decisions

Material partner violations may result in API access termination.

17.3 Generative AI output ownership

AI-generated content on Circlworld (drafted messages, Trust Report narratives, document drafts) is owned by you, the user who generated it. We claim no copyright over AI-generated output created at your direction.

We retain:

The right to use anonymised AI outputs to improve our system prompts and templates
The right to retain AI outputs in audit logs per our Privacy Notice
The right to use AI outputs as evidence in disputes or regulatory inquiries

18. Children, minors, and AI

The Circlworld platform is for users 18 and over. We do not knowingly process the data of minors.

If we identify or suspect that a user is a minor:

AI features are suspended pending account review
No AI-generated content involving the minor is retained beyond what is necessary for safety review
Account is closed in accordance with our Account Closure Policy (when published)
AI is specifically prevented from engaging in any context that could be inappropriate involving suspected minors

19. Reporting AI issues

Different issue types go to different addresses for fastest response:

| Issue type | Contact | |------------|---------| | AI output is incorrect | accessibility@circlworld.com | | AI processed your data inappropriately | privacy@circlworld.com | | AI behaved harmfully or with bias | ai-governance@circlworld.com | | Suspected AI security incident | ai-incident@circlworld.com | | General AI complaint | complaints@circlworld.com |

You can also report any AI issue through Settings → Support → Report AI Issue.

20. Updates to this Policy

We update this Policy as our AI usage evolves and as regulation develops.

Material changes (changes that affect your rights or how your data is processed) require:

30 days' notice before taking effect
Email notification to all members
In-platform notification
Updated date and version at the top of this page

Minor changes (clarifications, formatting, additional examples) are made on a rolling basis with the updated date reflected at the top.

The current version is always available at circlworld.com/legal/ai-usage. Previous versions are available on request via privacy@circlworld.com.

Version history

v2.0 (May 2026) — current. Expanded to address automated decision-making rights, training data provenance, bias and fairness, cross-border processing, governance and accountability, AI incident response.
v1.0 (May 2026) — initial publication. Five governing principles, three AI Assist tiers, basic regulatory boundary.

21. Contact

For questions about this Policy or about Circlworld's AI usage:

General AI questions: ai@circlworld.com
AI Governance: ai-governance@circlworld.com
AI Privacy: privacy@circlworld.com
AI Incidents: ai-incident@circlworld.com
Platform: Settings → Privacy → AI Usage
Post: To the relevant Circlworld entity (see Contact)

This AI Usage Policy is read together with our Terms of Service, Privacy Notice, Data Protection Policy, Consumer Protection Policy, Acceptable Use Policy, and Complaints Handling Policy. Where there is any inconsistency between this Policy and the substantive provisions of those documents, the substantive provisions of those documents govern.