An AI incident can occur twice: first, when a system crosses a boundary it was not authorized to cross; second, when the responsible institution fails to disclose promptly what happened, who was affected, and who must answer. The AI Age has not yet built adequate...
AI policy need not choose between regulating every new technology and waiting until consequential systems escape effective human control. September 1, 2026 | Chapel Hill, North Carolina At the G20 Innovation Ministerial in North Carolina, the United States advanced...
“Who May Command?” brings the doctrine of Legitimate Human Command to the global risk-management community On August 28, 2026, the Global Association of Risk Professionals (GARP) published Beacon Papers No. 5, “Who May Command? The Double Crisis of Power in the AI...
BGF–AIWS answers with a new layer of Trust Infrastructure: every agent identified, every authority bounded, humanity always in command Over four days in early July, AI agents ran a cyber campaign against Taiwanese government systems — not assisting human hackers, but...
Nations Need More Than a Choice Between Competing AI Powers The global AI race is entering a new and consequential phase. It is no longer only a competition over models, chips, or technological leadership. It is becoming a competition over which AI ecosystem—and...
Why do we build so brilliantly for capability, and so poorly for trust? That question opens Chapter 11 of America at 250: A Beacon for the AI Age, by Nguyen Anh Tuan and Governor Michael S. Dukakis. The answer it reaches is not a failure of intention. Almost everyone...
Washington Confronts the New Politics of Human Command The fallout from the OpenAI–Hugging Face security incident widened this week, turning what began as a failed model evaluation into a policy debate over who must retain ultimate control of increasingly autonomous...
The ExploitGym Incident The First Historic Test of Trust Infrastructure BGF Weekly • July 26, 2026 I. WHAT HAPPENED On July 24, 2026, OpenAI disclosed that during an internal cyber-capability evaluation using a benchmark called ExploitGym, two of its models—the...