Helping Save Lives

Any AI solution used in Laerdal products or services must, either directly or indirectly, support the overall company mission of helping save lives.

Transparency

Laerdal is transparent about the use of AI in customer facing products and provide both users as well as customers and partners with needed information, including the following:

• Samaritan: All products and services by Laerdal empowered by advanced data processing or AI, should be equipped with the Samaritan icon to keep users informed.
• System overview: Collaborators and stakeholders are provided with system architecture diagrams to establish the level of insights into the AI solution that is needed.

Human Role

Laerdal is mindful about the human role in relation to the development and use of AI and recognizes that in the end, humans have the responsibility:

• AI for human support: When AI solutions is assisting humans in some task, the AI solution must function as intended and humans behind the development and testing are responsible for the quality. However, the human being assisted still has the responsibility for the actions taken based on the AI solution.
• AI for automation: When AI solutions are applied in a way which is automating a task and there is no human directly involved, special attention to quality and analysis of consequences of unintended events is necessary. Furthermore, frequent reviews and adjustments must be conducted, since ultimately, humans are accountable for the AI solutions.
• Human agency: When possible given the purpose of the AI solution, Laerdal aim to provide the user with reasonable options for control and involvement, e.g. being able to pause, opt-out, undo and dismiss aspects of the AI solution.

Tests to Ensure Quality

All AI-solutions by Laerdal are tested to a degree which is proportional to the likelihood and consequences of an unintended incident. This includes:

• Software testing: Traditional testing as with any software, including automated and internal human-based testing as well as both prior to first release and regression tests for subsequent releases.
• User testing: Testing by real users to test assumptions of the development and avoid blind spots.
• Testing a trained model: If a custom machine learning model is trained as part of the AI solution, the quality must be assessed with machine learning specific assessments on test data set by data partitioning to avoid misleading positive performance metrics due to overfitting.  
• Incident projection: As an addition to the issues discovered in testing, the AI solution must be reviewed and the most relevant incident risks be identified.
  Post-release improvements: Laerdal strive to improve products continuously and aims to collect user reactions and test results also after initial release of a product.

Operations Governance

All AI solutions by Laerdal deployed for production which are impacting users, must be monitored and traced to enable robust operations and swift troubleshooting. This includes:

• Versioning: It must always be possible with precision to identify the version of each component in an AI solution. This entail that the digital solution is under version control.
• Logging: There must always be a log of the activities for an AI solution to be able to troubleshoot, investigate unintended incidents and ensure accountability.
• Monitoring: An AI solution should be monitored during operations to ensure that the service is up-and-running, is operating as intended and that AI performance over time is not deteriorating. The monitoring should include alerts to notify about unwanted or unusual events.
• Accountability: Laerdal have clear roles and accountabilities during development, release and operations of AI-powered products and services, and users should have a possibility to provide feedback and report incidents.

Data Protection

All development and operations of customer facing products, as well as internal solutions, must be conducted in a way which is upholding the trust that users, customers and partners have in Laerdal regarding protection of data. This includes:

• Data-safe service: Any use of AI, customer facing or internal, must be conducted in a way which is not sharing data in an unsafe environment. Use of external services must follow existing processes for approval before use.
• Legal standards: Any use of AI by Laerdal must of course be compliant with legal regulation and relevant standards – see subsection about “Guidelines for Compliance”.

Technology Vendors

Laerdal is using different technology vendors for the development of the AI solutions, including, but not limited to, the major cloud providers. For all technology vendors, Laerdal is committed to ensure:

• Sourcing: Technology providers needs to be sourced according to existing company processes, evaluating both the software and the supplier.
• Legal and standards: Any use of AI by Laerdal must of course be compliant with legal regulation and relevant standards – see subsection about “Guidelines for Compliance”.

Equality

Equality and fairness are essential for Laerdal as it ensures that the technologies deliver equitable outcomes for all users, regardless of their background. To ensure this, any AI solution must be subject to ethical considerations, investigations, and features. Aspects include:

• Universal design: AI solutions aim for the universal design principle, that is, being designed for everyone, regardless of age, ability, or status. This includes attention to accessibility.
• Bias: In development and testing, Laerdal assess potential biases and implement necessary investigations and mitigations to address any identified issues.
• Option for feedback: Laerdal aims to make it possible for users to provide feedback if, in their opinion, the AI solution has issues with fairness.

Artificial Intelligence

All development and operations of AI solutions must meet requirements of relevant legislations and standards for the use of AI. To ensure that and facilitate the work:

• AI solution assessment: It is recommended that for any customer facing product or service with an AI solution, an assessment is made of the AI Solution forming the knowledge base for review and approval.
• Expertise: For customer facing products and services, an assessments regarding compliance aspects of AI are made in collaboration with legal experts or other relevant subject matter experts. For internal AI solutions, it is recommended to seek expertise if in doubt.

Data Privacy

All development and operations of AI solutions, must meet requirements of relevant legislations and standards on data privacy. To ensure that and facilitate the work:

• Privacy by design: All products and services must integrate privacy considerations into the development from the outset and thereby minimize privacy risk proactively.
• Contractual clarity: Products and services which are storing or processing data must do so in accordance with relevant contractual agreements between Laerdal and customers or customer facing partners as well as between Laerdal and technology vendors. To ensure that, there is established a clear overview of the contractual landscape.
• Privacy Impact Assessment: The impact on privacy for a given product or service is established by a Privacy Impact Assessment (PIA).
• Expertise: For customer facing products and services, assessments regarding compliance aspects of data privacy are made in collaboration with legal experts or other relevant subject matter experts. For internal AI solutions, it is recommended to seek expertise if in doubt.

Information Security

All products and services of Laerdal must ensure information security and protect for data leakage and security breaches. To ensure that and facilitate the process:

• Secure software development: All products and services must integrate security considerations into the development from the outset and thereby minimize security risk proactively.
• Expertise: Assistance on assessment of information security is recommended.

Carbon Neutral

In line with Laerdal’s sustainability objectives, the development of AI solutions should focus on minimizing environmental impact:

• CO2 in product development: Early evaluations of anticipated CO2 emissions of products and services must to be conducted and incorporated into assessments.
• Continuously lower CO2: Recognizing that different AI solutions have varying carbon footprints, Laerdal strives to adopt the most environmentally friendly options available.
• Reporting: Laerdal regularly reports CO2 emissions associated with our products and services in company’s Sustainability Updates.