Miro Assist Practices
At Miro, we create AI solutions that are based on transparent models, establish trust, are human-controlled, and promote fairness and equity. By understanding the specific actions we take to put these principles into practice, our customers can make the right choices for their organizations.
Transparency
Build transparent models that foster accountability.
Miro neither trains models using customer data nor allows its partners to use customer data to train their models, without the customer’s explicit consent.
Miro does not retain data without the customer’s explicit consent. Data is solely used to generate a useful response with Miro Assist.
Miro leverages both external models hosted by partners such as Microsoft Azure OpenAI Service and in-house models hosted by Miro. All partnerships and models are clearly referenced for continuous transparency.
Miro does not persist any customer data related to AI-specific capabilities, nor does Miro cache data related to user prompts or generative AI functions, without the customer’s explicit consent.
Miro keeps its AI privacy policy updated so customers understand how data is used and why.
Miro asserts no rights to the content created by Miro Assist. Rights to the output generated by Miro Assist are determined by the underlying AI models themselves. For example, in the Microsoft Azure OpenAI terms of use, Microsoft Azure OpenAI assigns to customers all right, title, and interest to the output generated and returned by its services.
Trust
Establish trust by prioritizing privacy and security to safeguard customer data.
Customer content, including production data and backup data, is stored on servers in the EU to assist customers with their GDPR compliance. For more information, see EU Data Residency.
Miro partners with companies like Microsoft Azure OpenAI that offer industry-leading safety and security capabilities in single Miro tenant environments.
Data in Miro is encrypted in transit with the TLS 1.3 protocol and at rest with the AES-256 algorithm.
Miro conducts regular penetration tests, including OWASP Top 10 for Large Language Model Applications, to find and address any weaknesses in the platform ecosystem.
Miro continually vets, reviews, and manages vendor engagements and strategic partnerships in accordance with its policies and with applicable regulatory requirements.
Miro conducts tests to confirm separation of data between Miro customers.
Miro closely follows privacy and data protection developments in the AI space and strives to adapt as needed.
Miro and its customers are held accountable under the terms and conditions.
Human control
Allow customers to preserve control when using our AI-enabled features.
Miro Enterprise customers maintain administrative control when using AI-enabled features by electing to opt in or out of AI services.
Help Center articles and admin documentation provide useful tips and educational resources to help customers navigate AI in Miro.
User satisfaction surveys allow customers to share feedback, concerns, and opportunities for improvements, allowing Miro to assess the perceived value and impact of the AI solutions and to tailor them to meet customers’ needs.
Fairness & equity
Strive for inclusive solutions that promote fairness and equity.
Miro strives for its AI-generated content to promote fairness and equity. This is achieved by moderating training data and extensively evaluating the results of test data. This includes content moderation tools that filter out inappropriate, biased and/or problematic content.
Miro only works with AI models that are trained on reduced-biased datasets.
As Miro’s experience and innovation within the AI space deepens, and as the AI landscape changes, these principles and practices may evolve.