AI
The Withdrawal Notice Describes a Different Action
The June 30 letter did not just lift the Fable and Mythos controls. It wrote the shutdown back into the record as a software-export problem, leaving the access theory untouched.
AI Policy
Sci-Fi Saturday
The Seed and the Feed
Neal Stephenson imagined abundance with a root. The modern fight over open weights asks who gets to stand between the model and the person trying to use it.
AI
Who Gets to Use the Model Now?
The government switched Anthropic's model back on, but only for a list of approved names. It will not say what law the list runs on.
AI
Is an AI Answer an Export?
The first court test of Commerce's AI kill switch asks whether hosted model access exports technology, or just returns an answer.
AI
Buried the Lede: How Anthropic Drowned Its Own Safety Plan
Anthropic published a blueprint for how governments should switch off dangerous AI models the same morning it shipped one. Three days later the government switched off Anthropic's.
AI
Commerce Found the Kill Switch
Anthropic launched its most powerful public model on Tuesday. By Friday the government had used export-control authority to switch it off worldwide, and the legal basis is still unclear.
What-If
What If Companies Couldn't Use Humans as Liability Shields?
A car's crumple zone is steel that deforms on impact. In AI deployment, the crumple zone is the person who gets fired.
AI Policy
Promise Nothing, Claim Everything
A 59-question comparison of AI terms of service found six competing companies converged on the same answer: no warranties, no liability, and your data used for training.
The Signal
The Signal — March 27, 2026
Late edition — Claude went down overnight. Today: DoW's own records show it labeled Anthropic a supply chain risk for its "hostile manner through the press." A judge just blocked that. Plus: Sanders and AOC want to freeze AI data center construction.
Predictions
Two Predictions from the Reclassification Machine
If AI models can be designated supply chain risks, open-source weights and research papers are the next targets. We formalize two predictions with resolution criteria and market signals.
Predictions
When the Law Meets the Agent
Three predictions on agent testimony, impersonation law, and operational registration — scored through our six-axis PDM framework with timelines, milestones, and watch lists.
AI Policy
Building Moats, Not Bridges: How AI Became Bureaucracy's Best Friend
For £45, a website called Objector.ai will kill a housing project. The UK startup launched in late 2025 with a simple pitch: paste in a planning application, and the AI scans it for vulnerabilities. Within minutes, it generates an objection letter, ranks the arguments most likely to succeed, and