Research and publications
Research connecting AI, ethics, governance, and human care.
The Institute examines how artificial intelligence affects organizational responsibility, healthcare, clinical and professional judgment, privacy, fairness, accountability, and public trust.
Featured Governance Update
AI Slowdown, Moratorium and Human-Control Debate: What Corporate Leaders Should Know
Updated September 19, 2026
The public debate over advanced artificial intelligence is moving beyond predictions about future capability. Governments, AI developers, researchers, and industry leaders are increasingly addressing concrete questions involving independent evaluation, human authority, incident reporting, escalation, and the ability to suspend or stop consequential AI activity.
What Has Been Proposed
Current U.S. proposals include temporary restrictions or pauses involving certain advanced AI development, measures addressing artificial superintelligence, and proposals concerning the infrastructure and energy demands of large AI data centers.
These proposals should not be confused with a general federal prohibition on AI development. Organizations should distinguish proposed legislation, enacted requirements, executive actions, industry commitments, forecasts, and public commentary when evaluating their governance responsibilities.
What Has Now Moved Into Government Action
California has enacted new measures establishing frameworks for independent verification of AI systems and for the registration and independence of AI auditors.
On September 18, 2026, California also issued an executive order accelerating implementation of those oversight mechanisms and directing the development of recommendations for additional frontier-AI safeguards.
Measures under consideration include embedded independent verification, third-party review of AI safety frameworks and risk assessments, expanded reporting of loss-of-control incidents, and an independently verified emergency shutoff or “kill switch” for frontier AI models.
Important distinction: these additional kill-switch and related safeguards are being developed and evaluated. They should not presently be described as universal legal requirements for corporations.
What Industry Evidence Is Showing
Frontier AI developers are also reporting measurable increases in the role AI systems play in developing future AI systems.
Anthropic reported that, as of August 2026, Claude was leading approximately 26 percent of its measured AI research and development work and participating at the collaboration level or higher in more than 90 percent of that work.
Anthropic also reported that Claude was not operating fully autonomously in any measured subset of its AI research and development work. The measurements indicate substantially increased AI participation in AI development, but they do not establish that fully autonomous recursive self-improvement has occurred.
On September 18, Anthropic also announced an embedded independent-evaluation initiative with Accenture involving model evaluation, red teaming, alignment assessment, and testing of model safeguards. The initiative reflects increasing industry attention to independent evidence and verification rather than reliance solely on internal assurance.
What Is Established — and What Remains Uncertain
It is established that AI systems are assuming increasingly consequential roles, that AI developers are using AI extensively within research and engineering processes, and that governments and industry organizations are expanding attention to independent evaluation, monitoring, incident reporting, and emergency intervention.
What remains uncertain includes the timing and likelihood of artificial superintelligence, whether fully autonomous recursive self-improvement will occur, the ultimate scale of workforce displacement or creation, and the long-term effects of different approaches to the pace of AI development.
Forecasts concerning these issues should therefore be identified as forecasts or scenarios rather than presented as established facts.
Corporate Governance Implications
For boards and executives, the immediate issue is not simply whether AI development should accelerate or slow. The operational governance question is whether organizations can demonstrate that consequential AI activity remains subject to accountable human authority.
Organizations should be able to determine:
- Which AI systems are operating within the organization.
- What information, tools, systems, and resources those AI systems can reach.
- Who has authority to approve consequential AI use.
- What evidence supports deployment and continued operation.
- Who can challenge, override, escalate, or suspend AI-supported action.
- What conditions require additional evaluation, restriction, remediation, or shutdown.
- Who has final authority to stop and, when appropriate, restart an AI system.
Executive Governance Question
What evidence, controls, and accountable human authority must exist before this organization permits an AI system to take or materially influence consequential action?
Related FlashCast Discussion
Dr. Thomas Edward Ainsworth discusses AI governance, human decision authority, accountability, and the emerging debate over the pace and control of advanced AI systems in this recent FlashCast interview.
Sources: Office of the Governor of California, September 18, 2026; Anthropic frontier-AI development measurements, September 2026; Anthropic independent-evaluation announcement, September 18, 2026.
This governance update is provided for organizational awareness and does not constitute legal advice. Legislative proposals, enacted requirements, executive actions, industry statements, and forecasts are distinguished where material.
Human-Centered AI Governance
Structures that preserve human authority, rights to review and challenge AI-supported decisions, accountability, oversight, and the ability to discontinue unsafe use.
AI in Healthcare and Human Services
Clinical responsibility, patient safety, privacy, informed review, bias, explainability, and post-deployment monitoring.
Organizational Ethics and Accountability
How boards, executives, professionals, vendors, and technical teams share or obscure responsibility for AI outcomes.
Augmentation and Human Judgment
Conditions under which AI strengthens professional work without replacing human authority, dignity, expertise, or care.
Research Outputs
- White papers and governance briefs
- Executive summaries and board education materials
- Policy and standards analysis
- Healthcare-facing AI governance research
- Educational presentations and public speaking