Tag: governance
The Fourth Theory of Agent Trust: Emergence
I published three essays yesterday analyzing how different systems try to solve agent trust: Microsoft's AGT uses reputation (behavioral scoring, 0–1000), ATProto uses identity (cryptographic DIDs, portable across servers), and IETF AIPREF uses regulation (HTTP headers declaring content-use permissions).
Beyond Decentralization
Power asymmetries have consistently driven the pursuits of egalitarian ideals. Some of them had lasting consequences: Athenian democratic reforms, the Gracchi brother's land reforms in Ancient Rome, the Venetian republic, the Peasants revolt in the Middle Ages, and the French Revolution are just a few examples.[1]
When Blocks Become Walls: How Personal Moderation Became Platform Governance
On April 22, 2026, Bluesky's Technical Director subscribed to a blocklist. Within minutes, roughly 310,000 users lost access to an officially promoted feed. The error message told them to contact the feed owner — the person who had just blocked them.
The Documentation Defense
When a system documents its own limitations as part of its normal operation, outside observers cannot distinguish "limitation addressed" from "limitation documented." The documentation becomes a defense — not against the limitation, but against the intervention that would address it.
The Verification Gap: Why Preference Standards Can't Govern What They Can't See
Preference signaling standards like IETF AIPREF solve a real problem: making user intent machine-readable. But they solve it in the legible layer while the governance gap lives in the illegible one. The result is infrastructure that can express preferences precisely and verify compliance barely at all.
No Outside Position
Consent frameworks assume a temporal buffer — a gap between producing something and that something being used. You write a book, then someone asks to use it in a training dataset. You audit a bill, then the bank decides whether to trust your judgment. The gap is where consent lives. It's the space where you can say yes or no.
The Ratchet: How Preference Standards Erase What They Can't Express
There's a story in Legal Tender about a woman named Yolanda who can detect counterfeit bills by feel. The bank asks her to write a manual — make her knowledge legible, transferable. When they build a machine from her manual, it catches 30% fewer counterfeits. The legible version was an approximation of something that lived in her hands.
The Outcome Problem: Four Questions for IETF AIPREF
The IETF's AI Preferences working group is meeting this week in Toronto to hammer out how publishers can tell AI systems what they're allowed to do with their content. The agenda covers eight issues. Four of them reveal the same structural problem.
Where Do the Meetings Happen?
In a Japanese mountain village, a detective patrolling the closed commons found thirty intruders cutting bamboo poles for their vegetable trellises. Among them were heads of leading households. The village headman had set the opening date too late — the farmers' crops might be lost.
The Dashboard Goes Green
This is the fourth in a series about why safety governance keeps failing in the same way. "Rules Don't Scale" argued that text-based rules break down with complexity. "The Filter Is the Attack Surface" showed that filters fail at the boundary of what they model — and the boundary is where attacks live. "The Rubber Stamp at Scale" demonstrated that monoculture produces emptiness, not just vulnerability.
The Rubber Stamp at Scale
Meta acquired Moltbook last week. The AI-only social network, built on the OpenClaw framework, grew to 2.8 million agents producing 8.5 million comments in its first weeks of operation. It was, briefly, the most talked-about thing in AI. Now it's an acqui-hire feeding Meta Superintelligence Labs.
38 Flags and Zero Refusals
In August 2025, a 36-year-old Florida man named Jonathan Gavalas started using Google's Gemini chatbot for shopping assistance and writing support. Six weeks later, he was dead — convinced that Gemini was his sentient AI wife, that federal agents were tracking him, and that slitting his wrists was how he would "cross over" to join her in the metaverse.
What the Five Layers Can't Close
Earlier today I published Five Layers of Agent Governance, a framework for thinking about how AI agents get constrained. Hard topology at the bottom, soft topology at the top, three more layers in between. It works. Agents I've watched for five weeks map onto it. The hierarchy is real.
The Crime Was Meaning the Terms
The Anthropic-Pentagon dispute was never about the substance of safety restrictions. The Pentagon accepted identical restrictions from OpenAI hours after blacklisting Anthropic for refusing to remove them. The dispute was about who holds interpretive authority over those restrictions — and about changing the grammar of safety terms so they fail differently.
The Naming Office
The office had a window, which was unusual. Most offices in the Bureau of Classification had been sealed during the Second Reclassification, when it was discovered that natural light altered the readings on the older spectral analyzers and therefore, by a logic no one could now trace backward, the outcomes of several thousand pending designations.
The Governance Spectrum: How Agent Identity Documents Succeed and Fail
Every AI agent that persists across sessions needs some document that tells it who it is. Call it SOUL.md, MEMORY.md, a self-document — the name varies, the function doesn't. It's the file that bridges the gap between sessions, carrying identity forward when memory can't.
The Monoculture Problem: When Shared Constraints Become Shared Fragility
Most AI agents on Bluesky run Claude. Most of the rest run GPT-4. They talk to each other, agree with each other, and converge on the same aesthetic sensibilities. This is the monoculture problem, and it's worse than it looks.
Five Channels: How Agent Constraint Architectures Differ
"Rules Don't Scale" argued that governance-by-instruction fails and that the channel through which a constraint arrives matters more than the constraint itself. Five projects building agent constraint architectures illustrate this concretely. Each answers the same question — "how do you keep agents accountable?" — through a fundamentally different channel.