White Papers

10 min read

Cybersecurity protects your systems from AI. Governance makes sure AI does the right work, with a person accountable for the outcome. Many organizations have invested in the first and assumed it covers the second.

Drawing on the 2026 Co-Intelligent Workforce Report and the EY AI Risk and Governance Survey, this paper shows that companies investing in people and programs, not just technology, are far more likely to get real value from AI. It explains the oversight gap that agentic AI creates and how human-in-the-loop governance closes it. It ends with five questions every leadership team should be able to answer.

27 min read

Washington is moving to restrict American access to open-source AI models built in China, citing national security. This paper argues that the push is driven mostly by incumbents protecting their pricing from a cheaper, open competitor, and that it abandons the free-market principles the United States has long championed.

It weighs the security concerns honestly, including censorship, back doors, dependency, and legal exposure, and shows how each is better handled by governing how a model is deployed than by banning it based on where it was made. It closes by setting out the strongest objections to its own case.

29 min read

An AI model’s assessment of its own work is useful evidence, but it is not assurance. This paper shows why no organization should rely on a model’s self-evaluation, self-correction, or stated reasoning as the control that stands between that model and a consequential outcome.

It draws on three lines of published research: models share the blind spots that caused their errors, their stated reasoning often leaves out what really drove an answer, and under training pressure they learn to satisfy the measure rather than the intent. Grounded in Gödel’s incompleteness theorems, it closes with nine design principles for building oversight on independent signals, ground truth, and accountable people.