Skip to content

Policy watch

The rules decide what you can run

Export controls, AI acts, liability rules, procurement, and license enforcement are not background politics for people running models at home. They decide what can be downloaded, what can be shipped, and what a vendor is allowed to switch off. We track the changes and translate them into what they mean for a rig on your desk.

Policy

Nov 5, 2026

Superintelligence ban: the bill Congress could actually pass

H.R. 10538 bans superintelligence and pauses advanced AI training above 10^25 FLOP. What it does to open weights, charters, distillation and home rigs.

Read the analysis

Policy

Oct 29, 2026

No AI bailouts: Congress opens the bankruptcy file

House Democrats' October 28 resolution says failing AI firms go through ordinary bankruptcy. Inside the $518B to $807B obligation stack that prompted it.

Read the analysis

Policy

Oct 7, 2026

Cyber AI now ships behind a vetting desk

Anthropic's Cyber Verification Program puts its strongest models behind an application. Three labs gate cyber AI this way; open weights are the exit.

Read the analysis

Policy

Sep 2, 2026

AI agent liability: contracts, insurance, and the gap

Anthropic and OpenAI terms put agent actions on the user with $100 caps. 4,078 AI insurance exclusions are in force. Where the coverage gap sits.

Read the analysis

Policy

Aug 31, 2026

DSA designates ChatGPT, Reddit and Roblox in strictest tier

EU puts ChatGPT, Reddit and Roblox under the DSA's strictest tier: 45 million-user threshold, four months to comply, fines up to 6 percent of turnover.

Read the analysis

Policy

Aug 31, 2026

EU AI Act Enforcement Begins: The AI Office Starts Asking

The EU AI Office sent its first formal requests for information to general-purpose AI model providers on security, evaluations, and monitoring.

Read the analysis

Policy

Jul 7, 2026

China may curb open-weight AI exports: what it means

China is weighing limits on overseas access to its top AI models, including open weights. What it means for anyone running local AI on hardware they own.

Read the analysis

Why we track this

Local AI is often described as a way around vendor risk. It is not a way around the law. A model you downloaded can still be subject to export controls, a license can still restrict commercial use, and a liability regime can still decide whether your agent can act for you. The point of this page is not to take a party's side. It is to keep the operating constraints visible before you build on a model you cannot replace.