Data Squared urges US to adopt AI with high accuracy focus

Data Squared (Data2), a small veteran-owned business, has submitted a set of recommendations to inform the federal government's AI policy. The recommendations emphasise the adoption of hallucination-resistant AI and propose measures to counter procurement hurdles that complicate business operations. The company's response addresses the core challenges in federal AI adoption, including data fragmentation, security, and the need for near-99% accuracy in mission-critical applications.

Their proposal suggests that graph-based AI could bolster the federal government's efforts in this domain. Jon Brewton, Founder and CEO of Data Squared, highlighted the challenges faced by federal agencies, stating, "Across the world, teams are focused on integrating AI to further their missions. Federal agencies face particular difficulties in ensuring that when they use AI, they are meeting the highest security standards that Americans deserve."

"Agencies need AI that works across siloed systems with transparency. Our recommendations to the White House AI Action Plan reflect our belief in America to lead the world in responsible AI development and deployment."

Among the key recommendations, Data2 has put forward establishing a Federal Intelligence Interoperability Program, which would include creating shared data standards and pilot initiatives that demonstrate how graph-based AI can improve accuracy and reduce costs. The company underscores the importance of mandating clear provenance tracking and explainable reasoning, particularly in high-stakes AI projects and adopting zero-trust architectures.

Data2 also calls for modernising procurement processes to ensure seamless integration of AI solutions with existing agency data and security standards. This includes using agile contracting vehicles to expedite pilot deployments and scale successful AI implementations. Additionally, the company advocates for investment in training and workforce development to enhance proficiency in knowledge graph fundamentals and data security practices.

The company's response underscores the need to prioritise high-impact use cases for AI deployment, suggesting that targeted applications, such as cybersecurity and fraud detection, could significantly enhance mission outcomes. They argue that showcasing real-world applications that achieve near-99% accuracy is critical, particularly in sectors requiring utmost precision.

Data Squared believes that addressing these recommendations is vital for advancing AI infrastructure within government sectors. The company emphasises that knowledge graphs provide a method for unifying disparate data while ensuring provenance and transparency, essential for explainable AI systems. They argue that their platform, reView, supports transparent reasoning, secure collaboration, and continuous learning, thus offering robust infrastructure for high-stakes government AI applications.

Arvind Jain, Glean Co-founder & CEO, provides additional insight into the challenges agentic AI faces: "For agentic AI, there's so much friction that comes with accessing data in the underlying environment… Success only works after you've solved for this problem." This aligns with Data2's viewpoint on the role of knowledge graphs in supporting advanced AI workflows by offering accurate, explainable, and traceable actions.

Data2 concludes its submission by affirming the necessity for robust policies that address data fragmentation and security needs in AI deployment, aiming for the US federal government to adopt a comprehensive AI Action Plan. Their reView platform is posited as a central component that could unite previously siloed data sources while meeting the high-accuracy requirements expected by federal agencies.