2026.29: Mainframes and Main Characters

The author argues that Europe's reliance on renting AI models from US companies creates significant sovereignty risks, as demonstrated when the US government recently ordered Anthropic to disable its Fable model for all foreign nationals. This essay examines how dependence on foreign-controlled AI infrastructure leaves European institutions vulnerable to sudden access changes, and why open-source alternatives from Europe or China are not viable solutions.

The article analyzes emerging policy threats to open-source AI models, particularly citing White House discussions about potential executive orders that could restrict open models above certain capability thresholds. It argues that open models lack the lobbying power of closed-source alternatives and face serious regulatory challenges that could create a permanent capability gap within the next 6 months.

This article examines Sam Altman's proposal to give the US government a 5% stake in OpenAI, which would then potentially be distributed to American households. It explores the rationale behind such wealth-sharing schemes—including compensating creators whose work trains AI and providing a safety net against job displacement—and calculates what each household might receive (~$320). It also analyzes the political appeal and potential benefits for OpenAI in building public trust and securing favorable treatment from the Trump administration.

Security researchers discovered a critical flaw in Microsoft's Secure Boot protection system that has existed for over a decade, allowing attackers to bypass the security feature using old, vulnerable 'shim' binaries that were never properly revoked. The vulnerability affects both Windows and Linux devices and enables malicious firmware installation that persists even after OS reinstallation. This reveals a fundamental failure in Microsoft's security certificate management process for boot-time protection mechanisms.

The May 2026 newsletter highlights major pricing changes in the AI industry, with OpenAI and Anthropic shifting to enterprise pricing that affects larger customers. Anthropic appears to be scaling rapidly with a $65B funding round and claims $47B in annualized run-rate revenue. The newsletter analyzes these developments and suggests that coding agents for enterprise are becoming the first proven business models for major AI labs.