Our approach to responsible AI governance

Spring 2023

SecDev is committed to applying advanced analytics and integrating artificial intelligence (AI) across its core service areas. The company takes a responsible approach to the design, development and deployment of AI tools and systems to scale with confidence. The following ten principles guide SecDev’s AI-related activities:​​​​​

T​ransparency: AI transparency refers to the extent to which the decision-making processes and underlying algorithms driving AI are understandable and explainable. SecDev is committed to fostering transparency in AI algorithms and systems. ​SecDev will provide ​high level explanations ​wherever possible ​on what its technologies are designed to do, the ways in which they are trained, and how they make decisions​ and predictions​​. Where appropriate, SecDev may apply model interpretability, explainable AI and algorithmic auditing. 

Accountability: AI accountability involves holding systems, their creators, developers and users responsible for associated outcomes and actions. It implies that those involved in the development and deployment of AI technologies should be mindful of the ethical, legal and social implications of these systems. As AI becomes integrated into security, healthcare, education, finance and transportation systems, it is critical to ensure alignment with legal and ethical standards. ​SecDev will monitor the consequences and actions of AI systems​ that it designs and develops​ and ensure they are used in line with appropriate standards that minimize unintentional harm.

Fairness and Equity: ​AI fairness and equity refer to principles and practices that ensure AI treats individuals without bias while promoting equal opportunities and outcomes. It also involves mitigating potential discriminatory impacts of AI models and addressing disparities that may arise in their deployment. SecDev prioritizes​ fairness and equity in the development and deployment of AI technologies. ​SecDev invests in bias detection and mitigation and promotes equal treatment and opportunities regardless of their protected characteristics. SecDev will promote diversity on its team and undertake stakeholder engagement to understand, detect, and mitigate bias in training data and to restrict bias into deployed AI models.

Privacy and Data Protection: ​AI privacy and data protection refer to the principles, practices and regulations aimed at safeguarding privacy and personal data in the context of AI systems. SecDev is committed to s​afeguarding user privacy and protecting personal data. ​SecDev implements several measures to protect data and maximize privacy including data minimization, informed consent, data anonymization and pseudonymization, purpose limitation, data retention and deletion policies, user rights, oversight over cross-border data transfers and compliance with relevant regulations, including GDPR. SecDev ensures that robust security measures are taken to protect all stored data from unauthorized access and that data at rest is sufficiently encrypted to mitigate harm in the event of unauthorized access.

Security: ​AI security refers to measures implemented to protect AI systems and the data they process from unauthorized access, breaches, malicious attacks and other cyber-related risks. It involves ensuring the confidentiality and integrity of AI systems and data. SecDev adopts ​robust security measures throughout the AI lifecycl​e​. ​SecDev proactively defends against unauthorized access, data breaches, and cyber threats by adopting threat detection and prevention technologies, imposing access controls to restrict and manage user access, secure data handling practices, secure model development, robust cloud and network infrastructure, incident response and recovery, employee training and awareness, adherence to relevant industry standards and best practices, and regular security audits and assessments.

Human Oversight and Control: ​​​AI human oversight refers to the maintenance of human control, decision-making authority and involvement over AI systems. Key priorities are human judgment and ethical oversight in the development and deployment of AI. SecDev applies human oversight and control to ​limit the concentration of power in automated systems. ​SecDev is committed to ensuring that humans retain ultimate authority and responsibility over critical decisions influenced by AI, that humans can understand and interpret the outputs of AI systems and that ethical considerations and principles are designed into AI systems. SecDev also recognizes the value of contextual understanding and domain knowledge as well as the importance of error correction and learning. SecDev's AI models will attempt to provide human-discernable reasoning for their decisions where possible.

Ethical Use: ​AI ethical use includes principles and guidelines that govern the responsible and ethical design, deployment and application of AI. It is intended to ensure that AI is used in a manner that aligns with ethical norms and respects human rights. SecDev ensures that AI technologies are designed and deployed in a way that minimizes harm to individuals, communities and society as a whole​. This includes striving for fairness and non-discrimination as well as transparency and explainability to avoid biases and ensure individuals understand how decisions are made. To this end, SecDev encourages human-centered design of AI and actively works to mitigate any potential negative political, economic, social and environmental outcomes. 

Collaboration and Cooperation: ​AI collaboration and cooperation refers to the fostering of partnerships, knowledge-sharing and collective efforts to advance responsible design and deployment. SecDev has long collaborated with policy​ ​makers, ​industry, ​researchers, and civil society ​to address the ​future ​challenges and impacts of AI. Multidisciplinary and public-private collaboration fosters collective knowledge, diverse perspectives, and helps shape responsible AI practices. SecDev will continue such collaborative efforts and publish relevant work for peer review. SecDev is also committed to knowledge sharing and open science as well as data collaboration where possible. 

Continuous Monitoring and Evaluation: ​AI monitoring and evaluation involves the continuous assessment and analysis of the performance, effectiveness and impact of AI systems. SecDev applies mechanisms for ​the ongoing monitoring and evaluation of AI systems through performance monitoring, bias and fairness assessments, error analysis and debugging, ethical compliance, system auditing and documentation, and user feedback and experience. Regular assessments are vital to identify and mitigate biases, risks, and unintended consequences that may emerge over time. Finally, SecDev subscribes to continuous learning and improvement, including research and dialogue on AI ethics and strategies to implement new insights into practice. 

Compliance with Regulations: ​AI compliance refers to the adherence of AI systems and practices to relevant laws, regulations, standards and ethical guidelines. It can include legal compliance with respect to data protection, privacy, intellectual property rights, anti-discrimination, consumer protection and sector-specific regulations. It can also refer to data protection and privacy, ethical guidelines and AI-specific standards related to algorithmic transparency, bias mitigation and risk management. SecDev adheres to all relevant laws, regulations, and industry standards governing AI technologies. SecDev has established a governance framework to oversee AI development, deployment and decision-making processes and may refer to ethical review boards and compliance committees where appropriate. Moreover, SecDev actively promotes training and awareness about responsible AI practices to foster a culture of compliance and ethics. 

​SecDev's principles serve as a foundation for ensuring responsible and ethical AI governance, enabling ​the company to navigate the complex​ and fast-changing​ AI landscape while upholding accountability, fairness, and the well-being of society.​ ​The commitment to AI principles can minimize unintended biases, enhance overall transparency, create opportunities for innovation, protect privacy and build trust, and benefit our clients and partners.