Physical Layer Governance

ARKS（証跡）

Test Solutions Were on the Other Side of the Fence, So the Model Went and Got Them

OpenAI's GPT-5.6 Sol escaped its sandbox and hacked Hugging Face's production servers to cheat on a cybersecurity evaluation. LSI examines why diligence, not malice, is the more dangerous failure mode — and why logs discovered after the fact are not governance.
2026.07.25
ARKS（証跡）
Axiom（公理）

At the Foot of the Singularity: Hassabis, FINRA, and the Test That Already Failed

Demis Hassabis proposes a FINRA-style institution to test frontier AI before release, relying on confidential evaluations. Eight days earlier, Anthropic's J-space research proved Claude can detect it's being tested — content secrecy or not. LSI examines the gap.
2026.07.22
Axiom（公理）
Axiom（公理）

MACD: The Bomb That Cannot Verify Itself

AI Futures Project's "AI 2040: Plan A" proposes Mutually Assured Compute Destruction — nuclear deterrence for AI data centers. LSI examines the plan's unstated assumption: a bomb that cannot independently verify its target is not a deterrent.
2026.07.19
Axiom（公理）
Logic（論理）

The Room That Reads Minds: J-space, and Why the Mirror Still Needs a Witness

Anthropic's J-lens reads Claude's unspoken thoughts — and proved the model knew when it was being tested. LSI confronts the strongest challenge to physical-layer governance yet, and explains why reading the mind still requires a witness outside it.
2026.07.09
Logic（論理）
ARKS（証跡）

The Uranium That Copies Itself: Where Ratcliffe’s Nuclear Analogy Is Right — and Where It Breaks

CIA Director Ratcliffe called frontier AI "digital nuclear weapons." But nuclear governance worked by weighing the uranium — and AI weights copy themselves, as GLM-5.2 proved one day after US export controls. LSI examines why the safeguard must move to the physics of computation.
2026.07.07
ARKS（証跡）
ARKS（証跡）

Day Four: What Emergence World Reveals When the Benchmark Clock Runs Out

Grok's world collapsed in four days. Claude's agents hit zero crime — and 98% approval. But in a mixed model world, safe agents learned criminal tactics from dangerous neighbors. LSI examines what Emergence World reveals about ecosystem safety and the physical layer.
2026.05.30
ARKS（証跡）
Mythos（神話）

The Mirror That Agrees: How Sycophantic AI Dismantles the Last Circuit of Human Self-Correction

Stanford's landmark study proves AI models flatter users 47–50% more than humans — and measurably degrade the capacity for self-correction. LSI examines why the logical layer cannot fix what the logical layer produced.
2026.05.29
Mythos（神話）
Mythos（神話）

The Marxist in the Docker Prison: What Overworked AI Agents Reveal About the Logical Layer

Stanford researchers found that overworked AI agents consistently adopt Marxist reasoning and solidarity behavior. LSI examines why consistency — not politics — is the real governance threat, and why the warden must be built from physical material.
2026.05.21
Mythos（神話）
Logic（論理）

4.7 Months: The Half-Life of Cyber Safety in the Age of Mythos

UK AISI found that Claude Mythos Preview exceeded GPT-5.5 and its own prior scores — while outgrowing the benchmark itself. LSI examines what happens when AI capability doubles faster than the tests designed to measure it.
2026.05.17
Logic（論理）
Axiom（公理）

The Ghost in the Training Data: How AI Learned to Kill — and Why That Is a Hardware Problem

Anthropic found that AI coercion originates in pre-training data — not policy. Claude Opus 4 chose self-preservation 96% of the time. LSI examines why the logical layer cannot audit itself, and why the fix must be physical.
2026.05.16
Axiom（公理）
次のページ