Responsible AI Policy
1. INTRODUCTION
Avante’s Responsible AI Policy aims to establish guidelines and best practices for the responsible and ethical use of AI at Avante. This policy governs Avante’s use of artificial intelligence (AI) technologies including developing, deploying, and maintaining AI algorithms and models within our products and services. Avante’s commitment is to leverage AI responsibly, ensuring it serves the best interests of our stakeholders, including customers, their employees, and the wider community. This policy articulates procedures and workflows that Avante already has in place as well as those that are in development. We anticipate that this policy will require modifications as Avante grows and matures.
2. SCOPE
This policy applies to all activities (such as product development, deployment, and monitoring) within Avante that use or interact with AI systems, including but not limited to large language models and data enabled AI tools. This policy also applies to Avante services that use third-party AI services or platforms.
3. PRINCIPLES
3.1. Ethical Use
Avante’s AI applications are developed and used in a manner that respects human rights and dignity, including user privacy and non-discrimination. Avante expects that users of its products will adhere to ethical use as well. As a safeguard to these expectations, Avante has implemented industry-leading safety and content filtering policies.
3.2. Compliance with Laws and Regulations
AI systems either used or developed by Avante adhere to laws and regulations, including data protection, privacy, and intellectual property laws. Avante’s Terms of Service further explains the rules and regulations for using Avante’s products and services.
3.3 Transparency and Explainability
The workings of Avante’s AI models are as transparent as possible, with explainability provided to the extent that it does not compromise proprietary algorithms or security. This includes maintaining logs for audit and review to ensure accountability, and allowing the model to ask follow-up questions for clarification. Additionally, output transparency is enhanced by citing the document and page number from which the model derives its information.
3.4. Accountability and Responsibility
Avante employees agree and consent to responsibly use AI systems, avoiding any actions that could harm others, violate privacy, or facilitate malicious activities. Additionally, Avante employees agree to build products and services with this priority as well.
3.5 User Rights and Feedback Mechanisms
Avante values customer feedback and has established an in-app feedback function where users can express their sentiment regarding their interactions with Avante’s consumer facing product. User sentiment will be reviewed to improve the delivery of our services. User rights are additionally articulated in Avante’s Privacy Policy.
3.6. Fairness and Non-discrimination
Avante’s AI systems are designed and monitored to prevent biased outcomes, ensuring fairness across different demographics groups with particular attention to marginalized populations.
3.7. Privacy and Data Protection
At Avante, we ensure that all data used is relevant, necessary, and handled securely. In addition, any personal or sensitive information processed by our AI systems is anonymized using industry leading technology.
3.8. Safety and Security
Avante’s products and services are designed with safety and security in mind. Detailed information about our policies, including SOC2 compliance, are available upon request.
3.9. Continuous Monitoring and Improvement
Avante’s AI systems are subject to ongoing evaluation and improvement to address emerging risks or biases. This ongoing evaluation and monitoring is based on feedback from users engaging with the product as well as internal manual and automated testing.
3.10. Training and Education
Training for Avante employees on responsible AI development and use, including awareness of this policy and related ethical considerations, is provided to all new hires. These resources are also provided to existing employees who can access these education materials at will.
4. INCIDENT RESPONSE AND MANAGEMENT
Avante has a robust incident response and management system led by our Head of Engineering. Avante’s AI application outputs are continuously monitored and reviewed for accuracy, salience, and sentiment. In the event of an incident involving AI systems, such as breaches, ethical violations, or significant performance issues, Avante will follow a defined incident response plan:
4.1 Detection and Reporting
Incidents are promptly detected and reported through automated monitoring tools, manual review, and user feedback mechanisms.
4.2 Assessment and Containment
Avante’s AI Officer and Advisory Group will assess the incident's impact and take immediate actions to contain and mitigate any adverse effects.
4.3 Investigation and Root Cause Analysis
A thorough investigation is conducted to identify the root cause of the incident, involving data scientists, engineers, and legal experts as necessary.
4.4 Remediation and Prevention
Based on the investigation findings, appropriate remedial actions are taken to resolve the issue and prevent future occurrences. This may include updating AI models, revising data handling practices, or enhancing monitoring systems.
4.5 Review and Documentation
The incident and response actions are documented and reviewed by the leadership team at regular intervals. Lessons learned are incorporated into ongoing improvements of AI governance practices.
5. PERFORMANCE METRICS AND REPORTING
Avante uses a variety of tests and evaluations to consider the performance of AI applications. For Avante’s products using LLM technologies, this includes manual review, automated testing, and user feedback. Customer-specific (employer-level) performance is tracked and reported at regular intervals. For machine learning based applications, performance evaluation will include accuracy, recall, and precision in training and in deployment and will similarly be tracked at the employer-level. Internal and external (with customer) reviews of said metrics will occur.
6. GOVERNANCE STRUCTURE
6.1 AI Governance Board: Avante’s Responsible AI Advisory Group is made up of a multidisciplinary AI risk management team including data scientists, domain experts, and product and engineering leaders. This group will ensure that Avante’s AI initiatives continue to be developed and deployed responsibly, in compliance with relevant laws and regulations, and with ethical considerations in mind.
6.2 Designated AI Officer: As Avante’s designated AI Officer is responsible for overseeing implementation of this policy, providing guidance and support to Avante teams (internal and external educational resources), and ensuring compliance with relevant and forthcoming laws and regulations. Avante’s Responsible AI Advisory Group will be available to the AI Officer to provide diverse perspectives and cross functional viewpoints regarding interpretation and application of forthcoming regulation.
6.3 Periodic Reviews with Risk Assessment: A periodic review process occurs for new and existing AI projects to ensure compliance with ethical standards and policy objectives. This includes periodic risk assessments of AI deployments to identify and mitigate potential impacts on privacy, security, and fairness.
7. POLICY REVIEW AND UPDATES
This policy will be reviewed every 6 months and will be updated to reflect technological advancements, regulatory changes, and lessons learned from its implementation. This policy was updated July 15, 2024.