Making the world a safer place to work is our mission

Most of the AI and machine learning across our platform, from workplace monitoring to compliance checking to our in-app assistants, is built and trained by our own team.

Where we do bring in outside providers, we say so plainly: our AI assistants mainly run on Anthropic's Claude models, hosted inside our own cloud environment rather than sent to a public API. Two specific features use OpenAI (image analysis for Safety Agent, and course creation for Smart Courses). Licence plate recognition (when a site chooses to turn it on) uses a specialist computer vision model from Plate Recognizer, rather than a general-purpose AI provider.

No, each feature is built around a specific, named provider chosen for that job, rather than a customer-configurable option.

It reads your question in plain language, works out what you're asking for, and uses read-only tools we've built to look up the answer, then responds, or drafts something like an incident report or a form for you to check. It only ever looks in places you're already allowed to see (more on that in the Access section below).

Yes, like any AI, ours can occasionally get something wrong, especially with an ambiguous question or an unusual document. Where Ask Rapid can't find verified, trusted information to answer with, it's designed to say so rather than guess. Treat AI-generated answers, drafts, and alerts as a helpful starting point, not a final word, and always check anything AI produces before relying on it for a safety or compliance decision.

Yes, in whichever way makes sense for that feature. AI Simple Report sits inside a form you can always complete manually instead, if you'd rather skip the AI-assisted draft. Ask Rapid has its own On/Off switch per product. And Rapid Check's AI checks, Compliance Agent, Safety Agent, Plate Recognizer, and Rapid Monitor's face and object detection are all opt-in, none of them run on your site unless you've deliberately turned them on.

There is a caveat regarding RapidCheck. Every document uploaded to Rapid is processed centrally as part of our core document handling, regardless of whether RapidCheck's AI checks are switched on. Turning the feature off stops the AI-assessed results from being surfaced, but the underlying processing still happens and the data is still stored in our database.

No, though for RapidCheck, that's actually your choice to make. Auto-Approve is an optional setting you control per document type: switch it on for the checks you're confident in, leave it off where you want a person to make the final call, and every decision, automatic or manual, is logged and correctable by an admin at any time. Our Ask Rapid assistants can search for information and draft content, but a person always confirms before anything is submitted. AI Incident Reports only ever produces a draft incident report, you can edit it, dismiss it, or submit it, but nothing goes anywhere without you. Safety Agent doesn't act on anyone's behalf either: it only detects and raises alerts, using detection rules your own team configures, so even when it gets something wrong the result is a mistaken alert for a person to review, never an action taken without you.

Rapid Monitor's computer vision models are developed in-house and tested against real-world scenarios before release, with regular updates as conditions change. On the industry-standard benchmarks used to test facial recognition accuracy, our results are: LFW 99.83%, CFP-FF 99.87%, CFP-FP 94.63%, and AgeDB-30 97.91%.

No, we deliberately limit what AI features can see. Safety Agent, for example, only analyses the still images tied to a triggered alarm, never a continuous video feed. Our incident reporting assistant does not collect your location when helping you draft a report.

For most features, no. Anything running on Claude via AWS Bedrock (RapidCheck, Compliance Agent, Ask Rapid chat, and AI Incident Reports) and our proprietary Computer Vision models in use by Rapid Monitor stays within your region, Australia or Europe.

The two features that use OpenAI, Safety Agent's image analysis and Smart Courses are the exception: OpenAI's models are hosted in the USA. Plate Recogniser's data is also stored in the USA.

For all Rapid products other than Rapid Monitor, no. We only use your data to the extent necessary to provide the Services to you, not for other purposes, as stated in our Terms & Conditions. Our Privacy Policy goes further for biometric information specifically: Rapid Global does not use your biometric information for our own purposes, nor do we share it with or sell it to others. Any research and development we carry out is limited to collective learnings about how people use our services, like usage analytics, not building AI models on your data.

This is backed by our providers' own commitments, not just ours. Amazon Bedrock, which runs RapidCheck, Compliance Agent, Ask Rapid chat in Aware, Incident and Workforce and AI Simple Report, states that customer content is not used to improve the base models and is not shared with any model providers. OpenAI's API, used for Safety Agent and Smart Courses, carries the same default: business and API data isn't used to train their models unless a customer explicitly opts in, which we haven't. Plate Recognizer, used for licence plate recognition where enabled, processes plate images to return a read result under its own data handling terms.

Rapid Monitor works differently. As part of its implementation, Monitor uses footage from your site to train and tune custom detection models for optimal results. This data is not shared with third parties.

You can see every third party we work with, and what each one receives, in our Third-Party Providers list.

Yes, every AI-powered feature is covered by the same Rapid Global Terms & Conditions and Privacy Policy that already govern your account. There's no separate AI-specific contract to sign.

Every Rapid AI feature sits under the same standards as the rest of our platform: ISO 27001-certified infrastructure, and compliance with the Australian Privacy Principles and GDPR.

Yes. Surveillance footage lives on your own on-site server, and you control how long you keep it, but as a general rule, it's destroyed within eight weeks of capture. Face images we extract and store separately on the Rapid Global platform are destroyed within 90 days of collection, and any biometric information that isn't matched to an existing profile is deleted within 90 days too.

No. Every AI assistant in Rapid, whether it's answering a contractor's question, helping complete a form, or searching incident records, is limited to exactly what that person's account is already permitted to access. Nothing is shared across companies, contractors, or packages.

Yes. Every automatic decision, like a RapidCheck document approval, is logged and visible to admins, who can review or correct it at any time.

No. Our AI assistants can draft, search, and suggest, but creating or submitting anything (an incident report, a registration form) stays within your own permissions and requires your action.

For features like RapidCheck and Compliance Agent, the prompts sent to the AI are built entirely on our own servers, nothing a user types can reach or alter them. For assistants that do take free-text input, such as Ask Rapid, the AI can only use a fixed set of pre-built, read-only tools, so even a malicious prompt has nothing to act on beyond looking up information the person is already allowed to see. Changing what prompts our AI features can run is restricted to a small number of senior staff through a reviewed, permissioned process.

Every AI feature sits behind the same authentication and role-based permissions as the rest of the platform, none of them are open to anonymous or unauthorised traffic. We monitor for abnormal or malicious request patterns using the same tools that watch the rest of our platform, and several features add their own limits on top, like prompt-length checks and request validation.