AI safety and responsible AI

AI is an increasingly powerful technology that can benefit Canadian citizens, governments and industries. It has the potential to accelerate scientific discovery and create innovative solutions to Canada's greatest challenges. But it can also pose risks to its users and society.

For over 20 years, our Digital Technologies Research Centre has been at the forefront of responsible AI by developing and deploying AI systems in an ethical way. We are guided by principles of fairness, privacy protection, inclusiveness and sustainability to ensure our AI solutions align with human values and societal well-being.

We also have deep technical expertise in AI safety to identify and mitigate risks posed by advanced AI systems and ensure they behave as intended and avoid causing harm.

What we offer

AI safety research

We are a leading partner under the Canadian Artificial Intelligence Institute (CAISI), through which we conduct research supporting the Government of Canada's priorities on AI safety.

We work with CAISI and other partners to advance knowledge of AI safety and contribute to ensuring a safe adoption of AI in Canada and internationally.

Expert guidance on responsible AI

We work with partners and clients from industry, government and academia to provide guidance and inform policy on responsible AI. Our research teams can help businesses and government organizations ensure that they are deploying AI solutions responsibly, for example, by assessing their risks of bias, inaccuracies, data contamination and AI-specific cybersecurity vulnerabilities. Our experts can also make AI systems more transparent and more effective at human-AI interaction.

Developing responsible and safe AI-enabled systems

We have helped hundreds of clients develop and adopt responsible and safe AI systems and provided guidance to help identify issues and reduce AI risks. We can provide responsible and safe AI-enabled solutions that meet the planning and operational needs of government and industry clients.

Our programs

AI Accelerator

Helping Government of Canada departments and agencies harness the power of AI in a responsible way.

Data Analytics Centre

Ever wondered what your data is hiding? Our analytics services can unlock its potential.

Text analytics

Extracting information from text and image data.

Cybersecurity

Research to help prevent and address cyber threats.

Cryptography and quantum computing

New computing and cryptographic solutions for data security and privacy.

Multilingual text processing

Natural language processing, including machine translation.

Contact us

Yvan Gauthier

AI Safety

yvan.gauthier@nrc-cnrc.gc.ca

Stéphane Tremblay

Data Analytics Centre

stephane.tremblay@nrc-cnrc.gc.ca

Related links

Featured projects

Evaluating tools to detect AI-generated text

With the rapid advancements in AI, being able to detect whether a text was produced by a person or a machine has become essential for assessing the trustworthiness of the information.

This research project was carried out in partnership with Defence Research and Development Canada and involved a comprehensive review of existing approaches that allow for the detection of AI-generated text, including watermarking, statistical and stylistic analyses and machine learning classification.

The results of this research were published in 2025 in the Journal of Artificial Intelligence Research. They offer insight into the strengths and weaknesses of existing detection methods, provide practical guidance on choosing effective detection tools as well as possible directions for future research to help the AI safety community address security risks and combat the spread of misinformation.

International joint testing and evaluation of AI systems

Safety guardrails in AI systems may not behave equally across all languages, which represents a serious security risk.

Through the Canadian Artificial Intelligence Safety Institute, we're working with an international network of AI safety institutes to develop globally recognized methods for testing the safety and security of AI systems.

In a recent joint testing exercise, NRC researchers worked with experts from 8 other countries to evaluate autonomous AI systems that can interact with the world and make decisions, known as AI agents.

Our experts helped the international research team evaluate the effectiveness of AI at preventing harmful content in 9 different languages: English, French, Farsi, Hindi, Japanese, Kiswahili, Korean, Telugu and Mandarin. They also analyzed how well AI agents perform and behave safely across a range of cybersecurity tasks.

The findings from this work are helping to identify safety gaps in advanced AI systems and guide the development of stronger protections across all languages.

From: National Research Council Canada

 
 
Date modified: