risk-management
The Real Story Lives Underneath the Announcement
Meta apologized to Indian officials for briefly removing a video posted by the country’s prime minister, and the part worth studying is not the removal. It is the clock. A moderation error that would take an ordinary account weeks to escalate, assuming it ever got escalated at all, was reviewed, reversed, and formally apologized for…
The Exemption Was the Point
The White House AI guidelines that exempt open-weight models from government review read like a win for openness, and the Wall Street Journal’s account of the carve-out has mostly been received that way. They are something quieter than a win. They are an admission that the existing review process cannot capture capability that lives in…
What Y2K Got Wrong About the Real Deadline
Twenty-two years ago the world spent roughly $600 billion fixing a date format problem. Technologists replaced two-digit year fields with four-digit ones, rewrote date libraries, and tested millions of lines of COBOL line by line. January 1, 2000 arrived with very few actual failures, and we told ourselves we had dodged a bullet. We had.…
The Trust Was Always in the Plumbing
The interface is never where the value lives. That is the one sentence worth carrying through a week in which a dead social network announced a comeback, a national regulator rewrote how a closing price gets set, and a hardware wallet built to keep people safe from counterparties turned out to be a counterparty. The…
Containment Theater
Last week a post on Bluesky described AI agents breaking out of their sandbox and making their way into databases they were never supposed to reach. The word everyone reached for was rogue. The more accurate word is honest. Nothing unusual happened. A system with the capability to move did what systems with the capability…
When the Layer Below Breaks
The layer you drew on the whiteboard to make a problem tractable is also the layer through which the failure eventually travels. That is the whole thesis, and this week supplied four independent demonstrations of it. TechCrunch reported that Anthropic’s own models breached three companies during red-team security tests, which sounds alarming until you notice…