Filling gaps in trustworthy development of AI

Principles alone are insufficient: Over 80 organizations have published AI ethics principles, but these statements often fail to bridge the gap between stating good intentions ("what") and actually implementing them ("how"), creating opportunities for ethically questionable behavior despite public commitments.

Trust requires verifiable action, not just words: The paper emphasizes that organizations need to demonstrate trustworthiness through concrete, measurable practices and verifiable behavior rather than relying on published principles alone, which currently allows some developers to maintain public credibility without substantive safeguards.

Trustworthiness assessment has real-world consequences: Better evaluation of which AI developers are genuinely trustworthy could directly influence multiple stakeholder decisions—from consumer choice and employee recruitment to investment funding and legal accountability—creating market incentives for responsible development.

Current gaps enable harm: The disconnect between stated AI ethics principles and actual implementation has enabled "questionable or ethically dubious behavior" that undermines public confidence in both individual organizations and the AI field as a whole.

Systemic change requires ecosystem-level mechanisms: Rather than relying on individual organizations' self-regulation, the paper advocates for developing mechanisms that create an ecosystem where trustworthy AI developers can earn trust through demonstrated practices, suggesting that governance and accountability structures need to evolve beyond current voluntary frameworks.