Founding Principles of AI at Magnet Forensics

Our Founding Principles of AI serve as a living document, guiding how we build and integrate AI into our software. These principles ensure that AI enhances—but never attempts to replace—human judgment, while safeguarding privacy and maintaining the integrity of investigations. As our technology evolves, so too will this document, consistently providing our customers with a clear understanding of our approach to responsible AI development, and always in service of our mission to unlock the truth and protect the innocent.

1. Design AI to enhance—never replace—human judgment

• Principle: Our AI systems must be developed as tools that complement human investigators, speeding up processes while never undermining human oversight.
• Implementation: Design AI features to assist investigators in identifying leads, patterns, and connections. Ensure that all AI-generated insights are presented clearly for human verification, with mechanisms for investigators to confirm, modify, or reject AI suggestions whenever possible.
  • Focus on building AI that handles large data sets, highlights relevant patterns, and automates time-consuming tasks. However, create the system so that all AI-driven insights require human review and approval before any critical decisions are made. Implement safeguards that prevent the system from acting autonomously without investigator verification.
  • Build the system in such a way that investigators are always encouraged to critically evaluate AI outputs. Design interfaces that clearly present AI-generated information, but prompt users to verify or challenge the results before drawing any conclusions. Avoid building any functionality that could tempt investigators to rely solely on AI outputs.

2. Incorporate bias mitigation and objectivity from the ground up

• Principle: AI systems must be built to minimize bias and promote fairness, ensuring all individuals and evidence are always treated impartially.
• Implementation: Design the system to prevent users from excluding specific people or data without proper justification. Implement transparent processes that log and flag any attempts to exclude data or manipulate AI outputs, ensuring the system maintains impartiality throughout investigations to the greatest extent possible.
  Models should be evaluated for potential bias against different demographics including race, age, religion, or gender.

3. Prioritize transparency and explainability in AI systems

• Principle: AI-powered features must be transparent and provide accompanying reasons for all outputs, enabling investigators to trust and verify the system’s outputs.
• Implementation: Implement features and support mechanisms that give investigators context about how the AI reached a result. This may include noting the type of AI employed as well as annotated references, or “citations” to specific source data, when practical, for the investigator to assess accuracy of the AI output.

4. Ensure ethical use and data privacy protections

• Principle: AI must be developed with robust ethical safeguards and data privacy protections, operating strictly within legal requirements while empowering investigators to uphold ethical standards.
• Implementation: Incorporate mechanisms that enforce data privacy standards and prevent misuse of the AI system. Include features that prevent unlawful usage, unauthorized profiling, or other unethical actions. Ensure the system complies with all relevant privacy laws and regulations and implement clear auditing tools to track data use.

5. Maintain integrity of investigations with regular testing and updates

• Principle: AI models should be rigorously tested against well-curated, representative datasets to confirm their accuracy, repeatability, and robustness before being deployed.
• Implementation: Continuous performance monitoring is essential. Regular audits and re-assessments as well as periodic audits and model re-validations should be conducted to account for new and changing technologies.

6. Facilitate proper user training and education

• Principle: The effectiveness of AI depends on how well investigators understand and use it, so Magnet Forensics must ensure the system is intuitive and comes with educational resources.
• Implementation: Design the AI system with user-friendliness in mind, ensuring investigators can easily navigate, understand, and interact with AI-driven features. Provide built-in training modules, tutorials, and documentation to help users understand both the capabilities and limitations of the AI system, reducing the risk of misuse or over-reliance.
• Ongoing training: Implement regular updates to keep investigators informed about new features, improvements, and best practices, ensuring they remain aware of significant changes as the AI system evolves.

7. Design principles

• Principle: Design AI features to complement and empower forensic professionals, ensuring the AI never attempts to replace their expertise.

• Implementation: To gain users’ trust, the AI experience must be designed to:

• • Provide control to the user: Allow users to decide when AI is used and indicate which features are AI powered, wherever possible.
  • Clearly identify AI-generated content: Users should be able to easily recognize when information comes from AI.
  • Present understandable information: The data provided must be clear and easy to comprehend. Clearly outline how the AI arrived at its outputs, whenever possible.
  • Provide data verification: Offer links to original sources, enabling users to verify the information.
  • Encourage feedback: Allow users to confirm or reject AI-generated information to provide feedback. Ensure user feedback with AI plays a role in shaping future updates, reinforcing that AI is not just a tool, but a constantly evolving solution.

Additionally, AI features should:

• Streamline workflows: Help users quickly find relevant information, making the process more efficient.
• Support user needs: Offer suggestions and options to surface more relevant results when assistance is needed.

Magnet Forensics is dedicated to creating AI solutions that put investigators first. Our Founding Principles of AI embody this dedication to empowering forensic professionals and preserving the integrity of their work. By emphasizing fairness, explainability, privacy, and education, we aim to deliver tools that not only strengthen investigative processes but also earn and maintain the trust of our users. These principles will guide us as we continue to innovate, ensuring our AI remains a powerful ally in the pursuit of justice.