
AI Tools Data Policy Guide
For maritime professionals: what AI is doing in the tools you use, and how to manage it
Version 1.0 - April 2026
On this page
Overview
How to use this guide
Every tool below has AI capabilities either built in or added as a feature update. For each tool, we have linked directly to the vendor's data handling and privacy policy documentation - not their homepage. We have also included the specific steps to turn off model training or data sharing where an opt-out exists.
The question to ask of every tool on this list: Is this tool in use in our organisation, and do we know what it does with our data?
Section 1
The Major AI Models
These are the standalone AI assistants your staff may be using personally or for work, often without formal organisational approval or IT visibility.
Anthropic Claude
Team and Enterprise accounts are not used for model training.
API access is never used for model training.
On Team and Enterprise: training is off by default. No action required.
OpenAI ChatGPT
Team accounts have training off by default. Admins can configure data retention settings.
Enterprise and API access is never used for model training.
Alternatively, use Temporary Chat mode, these conversations are never used for training.
On Team: training is off by default. Confirm in your workspace admin settings.
Note: This is one of the few major platforms where model training is on by default for paid individual subscribers. If staff are using ChatGPT Plus for work, check whether this has been turned off.
Google Gemini (consumer)
Gemini Advanced accounts have stronger controls and the ability to turn off activity retention.
This stops conversations being saved and used to improve Google's models.
You can also delete your conversation history from this same settings panel.
Note: The consumer tier includes human review of conversations as a default. If Gemini is being used for work-related tasks on personal Google accounts, this warrants attention.
Policy links
Microsoft Copilot (consumer)
Copilot Pro subscribers can also manage settings within the Copilot interface.
Policy links
Perplexity
Free users have limited opt-out options, if sensitive research is being conducted, use Pro or a different platform with stronger controls.
Policy links
Section 2
AI Inside Tools You Were Already Using
These are platforms most organisations already subscribe to. AI features have been added, sometimes by default, sometimes via updates that were rolled out without a formal review. This is the invisible adoption category: AI that arrived inside tools already in use.
Microsoft 365 Copilot
(Word, Excel, PowerPoint, Outlook, Teams)
Copilot requires a paid Copilot add-on licence and is not active in standard Microsoft 365 plans.
Individual users can turn off specific Copilot features (such as meeting transcription summaries) within Teams settings.
Google Workspace: Gemini
(Gmail, Docs, Sheets, Slides, Meet)
Consumer Google accounts (personal Gmail) have different defaults, see the Gemini consumer entry in Section 1.
Individual users can turn off Gemini suggestions within Gmail by going to Settings → See all settings → General → Smart features and personalisation.
Adobe Acrobat AI Assistant
Adobe states that content submitted is not used to train Adobe Firefly models.
Administrators can disable AI Assistant across a team via Adobe Admin Console → Products → Acrobat → disable AI Assistant.
If your organisation processes sensitive documents through Acrobat (contracts, crew documents, charter agreements), confirm whether any staff have activated AI Assistant on those files.
Note: This is one of the most significant invisible adoption risks in maritime operations, Acrobat is near-universal and AI Assistant can be activated on any open document by any user with a subscription.
Adobe Creative Cloud: Firefly
(Photoshop, Illustrator, Premiere)
Enterprise agreements include stronger data isolation.
Enterprise admins can disable Firefly across a team via Adobe Admin Console.
Zoom AI Companion
Meeting summaries are stored in the account host's Zoom portal.
Admins can disable it entirely via Zoom Admin Portal → Account Settings → AI Companion.
Individual hosts can choose not to activate it per meeting.
All participants are notified when AI Companion is active, if you see a notification in a meeting, the host has it enabled.
Slack AI
Data stays within the Slack customer's environment.
Admins can disable it via Settings & Administration → Workspace Settings → Slack AI.
It is not active unless purchased and enabled.
Policy links
Notion AI
Notion uses third-party AI providers including OpenAI to power features. Content sent to Notion AI is processed by those providers under data processing agreements.
Individual users can avoid triggering AI by not using the AI commands in their pages, Notion AI does not process content passively.
Policy links
Canva Magic Studio / Magic Write
Enterprise plans offer stronger data controls and opt-outs.
Enterprise admins can configure team-wide AI data settings via the Enterprise admin panel.
Policy links
Dropbox Dash / Dropbox AI
Third-party AI providers are used under data processing agreements.
Dash is a separate subscription product that can be managed or cancelled independently.
HubSpot Breeze AI
Specific features such as content generation can be turned off independently from predictive scoring features.
Salesforce Einstein / Agentforce
The Einstein Trust Layer is designed to prevent customer data from being sent to third-party model providers.
Features can be turned off at the org level or configured per user profile.
LinkedIn AI Features
Direct link: linkedin.com/mypreferences/d/categories/data-for-generative-ai-improvement
Note: LinkedIn is one of the few major platforms that defaults to using your content for AI model training. If you post regularly, this means your professional writing is being used to train LinkedIn's AI unless you have opted out. Check this setting.
Policy links
Atlassian Intelligence
(Jira, Confluence)
Customer data is not used to train the base models.
It is not active by default on all plans, it requires a supported cloud plan and admin enablement.
GitHub Copilot
Telemetry data collection can be disabled separately.
To disable telemetry: go to GitHub Settings → Copilot → disable "Allow GitHub to use my code snippets for product improvements."
Organisations can restrict Copilot access via Organisation Settings → Copilot.
Policy links
Grammarly
Enterprise plans: customer data is not used to train AI models.
Enterprise admins can manage data retention and AI feature access via the Grammarly Business admin dashboard.
The browser extension can be disabled on specific sites by clicking the Grammarly icon → Settings → add the site to the exceptions list.
Policy links
DocuSign AI (IAM)
IAM uses Azure OpenAI under a data processing agreement.
IAM features can be managed via the DocuSign admin console. Standard eSignature customers are not affected.
Policy links
Box AI
Box AI uses approved AI providers under data processing agreements.
Admins can disable it via Admin Console → Apps → Box AI → disable.
Policy links
Asana AI (Asana Intelligence)
Third-party AI providers are used under data processing agreements.
Policy links
Monday.com AI
Third-party AI providers are used under contractual data protections.
Policy links
Meta AI
(WhatsApp Business, Messenger, Instagram)
WhatsApp Business API is subject to separate terms to the consumer app.
For WhatsApp Business API users, data handling is governed by the Business API terms rather than the consumer policy.
Review your WhatsApp Business settings and confirm with your API provider what data is processed.
Note: If your organisation uses WhatsApp for client, guest, or crew communication, which is common in yachting and maritime operations, Meta AI features may be active in those conversations. This warrants a specific review of your communications policy.
Apple Intelligence
(iPhone, iPad, Mac)
Requests requiring cloud processing use Apple's Private Cloud Compute, which Apple states does not store or use data for training.
When ChatGPT integration is invoked for complex Siri requests, that query is subject to OpenAI's terms.
Policy links
Xero
Xero does not sell customer data.
Analytics Plus predictive forecasting requires a separate add-on subscription and can be cancelled independently.
Policy links
QuickBooks (Intuit Assist)
Auto-categorisation is a core platform feature and cannot be fully disabled, but AI-generated suggestions can be reviewed and overridden before accepting.
Policy links
Mailchimp (Generative AI)
Subject line suggestions appear automatically and can be ignored. To disable them, go to Account Settings → AI features → manage preferences.
Otter.ai
Business and Enterprise plans offer stronger controls.
Business and Enterprise admins: manage data retention and training opt-outs via the admin dashboard.
Note: If Otter is connected to client, owner, or charter meetings, confirm whether recording consent has been obtained from all participants before the meeting begins.
Policy links
Fireflies.ai
Enterprise customers have dedicated data controls via admin settings.
Policy links
Fathom
Data is encrypted in transit and at rest.
It can be disconnected from your calendar at any time via Account Settings.
Policy links
Descript
Voice cloning requires explicit consent and can only be used on your own voice or with permission.
Voice cloning can be deleted via Account Settings → Voice Clone → delete.
Policy links
ElevenLabs
Voice data uploaded for cloning is subject to explicit consent terms at upload.
Account data deletion can be requested via their privacy team at [email protected].
Policy links
HeyGen
Avatar likeness data requires explicit consent at upload and is subject to strict use terms.
Contact HeyGen support to request full data deletion.
Policy links
Gamma
Third-party AI providers including OpenAI are used under data processing agreements.
For team accounts, admins can configure data handling in workspace settings.
Policy links
Section 3
Maritime-Specific Tools
The tools listed in this section are platforms built specifically for maritime operations. Each tool included here has published AI capabilities, we have only listed tools where the vendor has publicly stated their AI features, not where AI use is rumoured or unconfirmed. Where a vendor's AI data handling documentation is incomplete, this is flagged explicitly.
Yotspot
AI-generated summaries appear at the top of selected profiles to assist employers with candidate evaluation. Summaries carry a disclaimer to check for inaccuracies.
Policy links
MariApps OceanAI
(IntelliCrew, FaultSenseAI)
The privacy page provides a data erasure contact ([email protected]). Public documentation does not specify whether crew personal data is used to train MariApps' AI models.
A data processing agreement is available on request from MariApps and sets out the contractual terms for AI feature use.
Policy links
Martide
Policy links
Helm CONNECT
Privacy disclosures follow standard SaaS terms. Public documentation does not separately address AI feature data handling.
An AI-specific data handling addendum can be requested from Helm for organisations enabling API-based AI integrations.
Policy links
Veson Nautical / Q88: Shipfix AI
The notice follows GDPR structure. Public documentation does not include AI-specific model training carve-outs.
Shipfix AI has separate terms covering AI market intelligence features.
DNV Veracity
DNV states it does not use operational vessel data to train commercial AI models.
Data sharing with third-party applications via the Veracity marketplace is governed by separate API terms per integration.
LR OneOcean
(formerly ChartCo / Marine Press, acquired by Lloyd's Register 2022)
Public documentation does not include AI-specific model training carve-outs.
CyberOwl
Policy links
Windward AI
Policy links
Orca AI
Policy links
StormGeo
Policy links
Inmarsat Fleet Data / Fleet Xpress
(now part of Viasat, acquired 2023)
Acquisition note: Inmarsat was acquired by Viasat in 2023. Data handling practices are transitioning to Viasat's framework. The current data processing agreement reflects the Viasat terms.
Policy links
KVH Industries
Policy links
Closing
About this guide
This guide is a working reference, not a legal document. Every link directs to the vendor's own published documentation. Policies change. This guide reflects the position as of April 2026. Review the current version of each policy before making procurement or compliance decisions, and re-review annually or whenever a vendor announces a platform update.
If you need help auditing your organisation's current tool stack against these policies, Southern Sky AI's The Blueprint provides a structured assessment process. It starts by mapping what is already in use, including tools that may not have been formally reviewed, before any governance work begins.
Contribute
Suggest a tool or update a listing
For maritime professionals: If you use a tool that carries AI capabilities and it is not listed here, we want to know about it. Send the tool name, the vendor, and a link to their privacy documentation to [email protected] and we will review it for inclusion.
For maritime software vendors: If your platform has published AI capabilities and you would like to be considered for inclusion in this guide, please reach out to [email protected]. Inclusion is editorially assessed and based on published documentation. It is not paid placement. We require a direct link to your AI data handling terms before a listing can be added.
The Chart Room Dispatch
Weekly insights on AI in maritime, delivered every Sunday.
The twelve questions an AI policy has to answer are in the checklist.
Sign up for the DispatchSouthern Sky AI - Structured AI adoption for maritime
