AI & Trade Intelligence Notice
Gloseg B2B Limited Effective Date: August 16, 2026 Version: 1.0
Definitions
- "AI Feature" means any Gloseg functionality that uses machine learning or generative models to assist a User, including search and discovery ranking, HS code suggestion, translation, content summarisation, market intelligence summaries, drafting assistance, and support triage.
- "Output" means any suggestion, summary, translation, draft, classification, or ranking produced by an AI Feature.
- "Model Provider" means a third-party service that hosts or runs the underlying machine learning model used by an AI Feature.
- "HS Code" means a Harmonized System classification code used for customs and trade purposes.
- "Consequential Decision" means a decision with legal, financial or account-status effect on a User, such as suspension, listing removal, or payout action.
- "Prompt" means the input text or data submitted to an AI Feature to generate an Output.
1. Purpose and scope
1.1 What this notice covers
This notice explains how Gloseg uses AI across the Platform, the advisory nature of AI Output, the limits of AI accuracy, and User obligations when relying on AI-assisted features. It applies to every AI Feature on the Platform.
1.2 Who this notice applies to
1.2.1 All Users
- Every Organization and Member using an AI Feature, whether directly (asking a question) or indirectly (receiving an AI-ranked search result), is subject to this notice.
2. Where AI assists on Gloseg
2.1 Feature inventory
2.1.1 Search and discovery
- AI helps rank and surface relevant listings, suppliers, and buyers based on query intent and platform activity.
2.1.2 HS code suggestion
- AI suggests possible HS codes for a listed product based on the description provided, to help sellers and buyers start their own classification process.
2.1.3 Translation
- AI translates messages, listings, and forum posts to support cross-border communication.
2.1.4 Summarisation
- AI summarises long threads, market intelligence content, and news relevant to a User's sector.
2.1.5 Drafting assistance
- AI helps draft messages, listing descriptions, and RFQ text based on User input.
2.1.6 Support triage
- AI helps route and prioritise live chat and support tickets to the right queue or resource.
3. The advisory nature of every output
3.1 Not authoritative
3.1.1 Express statement
- Every AI Output is advisory. AI Output is not legal advice, customs advice, tax advice, sanctions advice, or investment advice, and must not be treated as such. Trade classification, sanctions determinations, and legal advice are never delivered by AI as authoritative on Gloseg.
3.2 Human review of consequential decisions
3.2.1 No solely automated consequential decisions
- Gloseg does not make Consequential Decisions about Users based solely on automated processing. Where an AI Feature flags a potential issue (for example, a suspected policy violation), a human reviews the matter before final action is taken.
4. Accuracy limitations and known failure modes
4.1 Why AI can be wrong
4.1.1 Hallucination
- AI models can generate plausible-sounding but incorrect statements, including invented facts, invented sources, or invented regulatory citations.
4.1.2 Stale data
- Market intelligence and summarisation features may reflect data that is out of date at the time of reading.
4.1.3 Translation drift
- Machine translation can shift meaning, especially for trade-specific or legal terminology, and should not be relied on for binding contractual language.
4.1.4 Misclassification
- HS code suggestions can be wrong, especially for products with ambiguous or composite characteristics, and incorrect classification can carry customs and legal consequences.
5. User obligations to verify independently
5.1 Independent verification required
5.1.1 What Users must check themselves
- Users must independently verify HS codes with a licensed customs broker or the relevant customs authority before relying on them for shipment or duty calculation.
- Users must independently verify applicable duties, taxes, and import or export licences with the relevant authority or a qualified adviser.
- Users must independently verify sanctions status of counterparties, goods, and jurisdictions through appropriate screening, rather than relying on the absence of an AI-generated warning.
5.2 No warranty of correctness
5.2.1 Disclaimer
- Gloseg makes no warranty that any AI Output is accurate, complete, or fit for a particular customs, tax, sanctions, or legal purpose.
6. Data used for AI features
6.1 What is used
6.1.1 Inputs
- AI Features may use listing content, RFQ and message text the User submits into the feature, and publicly available market data, to generate Output.
6.2 What is never sent to model providers
6.2.1 Exclusions
- Payment credentials, bank account and payout details, and government identity document numbers are never sent to a Model Provider as part of an AI Feature Prompt.
7. Confidentiality of RFQ and negotiation content
7.1 Handling of sensitive trade content
7.1.1 Confidentiality expectation
- RFQ and negotiation content used within an AI Feature (for example, drafting assistance for a quote response) is used only to generate the requested Output for that User's own conversation and is not shared with unrelated third parties or used to train models on behalf of other Users' benefit beyond the aggregate improvement of the Platform's own features.
8. Retention of prompts and outputs
8.1 Retention approach
8.1.1 Retention period
- Prompts and Outputs are retained for as long as necessary to provide the feature, support quality review, investigate misuse, and meet legal obligations, consistent with the Privacy Policy's retention principles.
9. Opt-out where a feature is optional
9.1 Which features can be turned off
9.1.1 Optional features
- Drafting assistance, translation suggestions, and market intelligence summaries can be turned off in account or content settings where offered as optional. Search ranking and support triage are core to Platform operation and cannot be fully disabled, though Users may still browse and search using manual filters.
10. Provenance and labelling of AI-assisted content
10.1 Labelling standard
10.1.1 Visible indication
- Content substantially generated or drafted by an AI Feature is labelled as AI-assisted where displayed to other Users, so recipients can apply appropriate judgement.
10.2 AI-drafted legal or compliance-adjacent drafts
10.2.1 Unpublished until reviewed
- Where an AI Feature drafts legal-adjacent content (for example, a compliance-style statement or a formal notice), the draft remains unpublished and marked as a draft until a human reviews and approves it before publication.
11. Market intelligence data sources
11.1 Third-party and indicative
11.1.1 Nature of the data
- Market intelligence pages draw on third-party and aggregated data sources that are indicative rather than guaranteed. Prices, demand trends, and forecasts shown are estimates for general orientation, not a warranted market position.
11.2 No guarantee of accuracy
11.2.1 Disclaimer
- Gloseg does not guarantee the accuracy of any price or forecast figure shown in market intelligence content and is not liable for business decisions made in reliance on it.
12. Prohibited uses of AI features
12.1 Misuse categories
12.1.1 Deceptive listings
- Using AI drafting assistance to generate misleading specifications, false certifications, or exaggerated claims in a listing is prohibited.
12.1.2 Evading moderation
- Using AI to reword content specifically to bypass moderation filters while preserving prohibited meaning is prohibited.
12.1.3 Mass scraping
- Using AI-assisted tools to orchestrate mass scraping of Platform content is prohibited and is treated as a breach of the Acceptable Use Policy.
13. Intellectual property in generated content
13.1 Ownership of drafting output
13.1.1 User ownership of their drafts
- Where a User directs an AI Feature to draft content for their own listing, message, or post, the resulting Output, once reviewed and adopted by the User, is treated as the User's own content for purposes of Platform ownership rules, subject to the User's compliance with this notice and the Acceptable Use Policy.
13.2 No exclusivity over underlying models
13.2.1 Shared infrastructure
- Users do not acquire rights in the underlying AI models, prompts engineering, or ranking algorithms used to power AI Features.
14. Incident reporting for harmful output
14.1 How to report a problem
14.1.1 Reporting channel
- Users who encounter an AI Output that is harmful, discriminatory, factually dangerous, or otherwise concerning should report it to legal@gloseg.com with the feature name, the Prompt (if shareable), and the Output received.
14.2 Gloseg's response
14.2.1 Review and correction
- Gloseg reviews reported incidents and may adjust, disable, or restrict the relevant AI Feature pending investigation.
15. Changes to this notice
15.1 Updates
15.1.1 Notification
- Gloseg may update this notice as AI Features evolve or as legal expectations for AI transparency develop. Material changes are notified through the Platform or by email at least 14 days before taking effect.
16. Governing law and jurisdiction
16.1 Governing law
16.1.1 Ghana law
- This notice is governed by the laws of Ghana, without prejudice to mandatory rights under GDPR, Nigeria's NDPR, or Ghana's Data Protection Act applicable to a given User.
16.2 Jurisdiction
16.2.1 Courts
- Subject to mandatory jurisdictional rules, the courts of Ghana have jurisdiction over disputes arising from this notice.
Contact
Legal notices, questions about this document, data protection requests and compliance enquiries: legal@gloseg.com
Data subject requests may also be submitted through the Privacy & My Data surface at /privacy/my-data.
Gloseg B2B Limited Accra, Ghana, West Africa
17. Illustrative scenarios
17.1 HS code suggestion scenario
17.1.1 Worked example
- A seller lists processed cocoa products and the AI Feature suggests a candidate HS code range based on the description entered. The seller must confirm the correct code with a licensed customs broker before shipment, since composite or processed goods often sit at classification boundaries that AI suggestions do not reliably resolve.
17.2 Translation scenario
17.2.1 Worked example
- A buyer in a Francophone market receives a translated version of a seller's message describing payment terms. Because translation drift can alter conditional language (for example, terms around partial shipment or penalty clauses), both parties should confirm final terms in a shared reference language or through a bilingual review before contractual commitment.
17.3 Market intelligence scenario
17.3.1 Worked example
- A market intelligence page shows an indicative price range for a commodity based on aggregated third-party sources. A buyer using this figure to set a target purchase price should treat it as a starting reference point, not a locked-in market price, and should validate current pricing directly with sellers.
18. Interaction with human support
18.1 Escalation from AI to human
18.1.1 Support triage handoff
- Where AI-assisted support triage identifies a complex or sensitive query (such as a dispute, sanctions concern, or account enforcement question), the query is routed to a human support agent rather than resolved solely by an automated response.
18.2 Requesting human review directly
18.2.1 User-initiated escalation
- A User dissatisfied with or unsure about an AI Output may request a human review at any time through live chat support or legal@gloseg.com, referencing the AI Feature and Output in question.
19. Training and improvement of AI features
19.1 How Gloseg improves features over time
19.1.1 Aggregate improvement
- Gloseg may use aggregated and de-identified interaction data to improve the relevance and accuracy of AI Features over time. Such improvement work does not involve sharing an individual User's confidential RFQ or negotiation content with unrelated third parties.
19.2 No resale of prompts
19.2.1 Restriction
- Gloseg does not sell User Prompts or Outputs to third parties for purposes unrelated to operating and improving the Platform.
20. Coordination with other policies
20.1 Reading together
20.1.1 Related documents
- This notice should be read together with the Acceptable Use Policy (for prohibited uses of AI Features), the Privacy Policy (for how personal data within Prompts and Outputs is protected), and the Cookie Policy (for how AI-related analytics may be measured on the Platform).