AI Assistant for Pesticide Scientific Literature Review

Research & Development · Enforcement

What it collects

Operational data

Anonymized data

Pesticide grey literature and published scientific literature (studies, reports, assessments) are ingested as the primary data sources. These are documentary, non-personal data sources used to support regulatory chemical review.

About behaviour

Anonymized data

Scientists can upload their own files to the system for processing. These user-uploaded files may include working documents related to pesticide reviews and are processed as part of the RAG pipeline.

Run by: Health Canada (HC)
Where: No fixed location
Kept: Retained not specified in source
Shared with: Accountable organization
Your copy: You cannot see the data it holds about you. What you can do

What it is for

Research & Development

Enforcement

Science-GPT is an AI-powered tool built by Health Canada's Pest Management Regulatory Agency (PMRA) to help scientists search, summarize, and extract information from scientific literature related to pesticide assessments. It uses large language models (LLMs) and retrieval-augmented generation (RAG) to speed up the review of grey literature, published studies, and scientific documents. This tool is used internally by Government of Canada employees only and does not involve personal information. Users are informed that AI is in use.

What it collects and what happens to it

Data taken in

Operational data

Anonymized data

Pesticide grey literature and published scientific literature (studies, reports, assessments) are ingested as the primary data sources. These are documentary, non-personal data sources used to support regulatory chemical review.

About behaviour

Anonymized data

Scientists can upload their own files to the system for processing. These user-uploaded files may include working documents related to pesticide reviews and are processed as part of the RAG pipeline.

Processing

Language Models

Large language models (LLMs) are used for summarization, entity extraction, and retrieval-augmented generation over pesticide-related scientific literature. The system is designed with subject matter experts (SMEs) to mitigate hallucination risks.

Search & Retrieval

Custom retrieval-augmented generation (RAG) pipeline enables scientists to search and retrieve relevant passages from grey literature, published scientific literature, and user-uploaded files related to pesticide chemical assessments.

What it does

Understanding (Semantic AI)

Human decides

The system performs search and retrieval, entity extraction, and summarization over scientific literature. Scientists query the system and review its outputs — all decisions about pesticide assessments remain with the human scientist.

Creating (Generative AI)

Human decides

The system uses LLMs and custom retrieval-augmented generation (RAG) to produce summaries and synthesized answers from scientific literature. Outputs are advisory; scientists review for hallucinations and accuracy before use in regulatory work.

Outputs

A recommendation or prediction

Anonymized data

The system outputs summaries, extracted entities, and retrieved passages from scientific literature that assist scientists in their pesticide assessment work. These are advisory outputs; regulatory decisions are made by human scientists.

Generated content

Anonymized data

The LLM components generate synthesized text summaries and answers from the retrieved scientific literature. Generated content may contain hallucinations; subject matter expert review is a built-in requirement before use in regulatory decisions.

Run by

Health Canada (HC)

Health Canada's PMRA is deploying Science-GPT to assist its scientists with pesticide chemical reviews. The project is supported by Health Canada Digital Transformation Branch (DTB), Shared Services Canada Science Experimentation Zone (SSC), and the Public Health Agency of Canada (PHAC).

Government of Canada AI Register — Science-GPT

Built by

Government of Canada

The system was developed by the Government of Canada, with no external vendor listed. Development involves Health Canada DTB, SSC, and PHAC as supporting partners.

Government of Canada AI Register — Science-GPT

Kept for

Retained not specified in source

The register does not specify a retention period for data processed by Science-GPT. As the system does not involve personal information, retention concerns primarily apply to scientific document corpora and query logs.
Duration: not specified in source

Shared with

Available to the accountable organization

Outputs and system interactions are accessible to Health Canada (PMRA) scientists and supporting government partners (DTB, SSC, PHAC). The system is restricted to GC employees only.

Not available to me

The system is an internal government tool available only to GC employees. Members of the public and affected parties do not have access to query outputs or system interactions.

Stored

Stored locally

The system is hosted in the Shared Services Canada Science Experimentation Zone (SSC SEZ), which is a Canadian government infrastructure environment, suggesting data is stored within Canadian jurisdiction.
Duration: not specified in source

How to read the colours

Can it identify you?

Anonymized data: Data about people with the link to who is broken. Stripped of identifiers, blurred, aggregated, or noised so this system can’t reasonably tie a record back to an individual.
Pseudonymous data: Each person’s data is tied to a token (hash, ID, template) that lets this system recognise the same person across events, but the token itself doesn’t reveal a name. Reidentification is possible with extra information.
Identifiable data: The data either contains a direct identifier (name, address, account name, recognisable face or voice, plate number) or carries a token this system uses to look up legal identity during processing.

Who completes the loop?

Human decides: This mode suggests; a person decides what to do next. The AI is always advisory — a human is in the loop on every decision. Example: a triage tool ranks cases for a clinician who chooses which to see first.
Human executes: This mode decides; a person carries out the result. Example: an optimizer plans the day’s trash-collection routes, and drivers run them.
Autonomous: This mode decides and acts on its own. No person reviews each decision or carries out the resulting action.

Definitions from the DTPR standard. Amber is about your data, violet about who decides. The fuller the shape and the deeper the colour, the more identifying the data or the less a person is involved.

What you can do

Ask about this system

Questions go to the Helpful Places, not the vendor.

Your rights

Right to Be Informed of AI UseAI use is disclosed to users of the system. GC employee scientists are informed that Science-GPT is an AI-powered tool when they use it.
Right to Algorithmic TransparencyThe system is registered in the Government of Canada's public AI register, providing transparency about its purpose, capabilities, data sources, and development status. The register entry is publicly accessible at the source URL provided.

Risks and safeguards

Reputational harmLLMs can generate hallucinations — plausible but inaccurate or fabricated content — which, if incorporated into pesticide regulatory assessments without review, could compromise the scientific integrity of regulatory decisions and harm the reputation of PMRA's assessment process.Safeguard: the system is built and tested with subject matter experts (SMEs); scientists are required to review all AI outputs for accuracy before use; AI use is disclosed to users.
Societal & cultural harmIncorrect AI-generated summaries or missed relevant studies in pesticide reviews could lead to flawed regulatory decisions affecting public health and the environment. The system is in development and is being validated before operational deployment; human scientists retain decision-making authority over all regulatory outcomes.