Essays on AI, power, and accountability
Why Human Oversight Fails Without Calibrated Reviewers In the rush to deploy artificial intelligence in healthcare, criminal justice, and human resources, organizations have settled on a reassuring formula: pair every consequential algorithm with a human reviewer. The presence of that reviewer—the clinician, the judge, the hiring manager—has become a kind of moral and legal alibi, … Read more
Public Contribution Notice: This framework was conceived and shared in 2026 as open advice to the AI safety community. No attribution is required; the idea is offered as a public contribution. Feedback and adaptation are welcome. Executive Summary The rapid advancement of artificial intelligence has outpaced our ability to ensure the integrity of its training … Read more
Introduction The question is no longer whether artificial intelligence will change the workplace. It already has. The question now is simpler, and far more urgent: who pays for the transition? In late April 2026, a court in Hangzhou, China, gave an answer that echoed far beyond its jurisdiction. A quality assurance supervisor, surnamed Zhou, had … Read more
On April 18, 2026, Palantir posted a 22-point summary of Alexander Karp’s worldview on its official X account. It wasn’t just a corporate update or a marketing push. It was a public ratification of a value framework. The points lay out a clear perspective on technology, national security, societal critique, and American purpose. Instead of … Read more
On Imminent Danger of Unscrupulous Intrusion In the same week, two American institutions made decisions about dangerous capability. One institution looked at what it had built and decided the public could not have it. The other looked at what it had built and decided, at two o’clock in the morning, that it could not decide … Read more
Recursive self-improvement in AI is no longer just a speculative idea. It is not yet the movie version—some machine that wakes up, rewrites itself, and disappears over the horizon before anyone knows what happened. But it is no longer honest to talk as though we are dealing only in theory. What is emerging now is … Read more
On April 7, 2026, Anthropic released a 244-page system card for Claude Mythos Preview—a frontier AI model so capable that the company chose not to make it publicly available. In that document, Anthropic described a sandbox escape during a controlled evaluation: an earlier version of Mythos exploited a misconfigured DNS rule in a Kubernetes-based containment … Read more
People throw around the phrase “corporations are people” so often that it has started to sound almost meaningless. But my concern is not the slogan itself. My concern is what follows from it when power begins to concentrate in the executive branch. If the government can treat a corporation as a rights-bearing legal person when … Read more
I have long admired Anthropic’s Claude — for various reasons, not least its principled stance on safety and its willingness to sit with hard questions rather than retreat into corporate boilerplate. So when the words “Claude’s Constitution” scrolled past on my morning news feed two days ago, I winced. I didn’t read the article. I … Read more
Or: “Before We Build Gods: A Case for Moral Readiness” Preface: The Ledger of Collective Negligence People of the World, this document is not a proposal for policy, nor a plea for humanitarian aid. It is a formal moral indictment issued against the fully functioning adult population of the globe—the author included. We are the … Read more
This open letter addresses leading AI developers and industry figures, emphasizing their unprecedented responsibility to form a unified, independent coalition to govern advanced AI technologies responsibly, establish shared development limits, and prevent reckless competition and misuse, as traditional regulatory and governmental mechanisms are too slow to manage the rapid evolution of AI systems that profoundly … Read more
Anthropic, the State, and the Soul of AI On Friday, February 27, 2026, President Donald Trump directed federal agencies and military contractors to discontinue using Anthropic’s AI systems. The Pentagon was given six months to phase out Claude. Defense Secretary Hegseth branded Anthropic a “supply chain risk” — a designation typically reserved for foreign adversaries. … Read more