Last Week in AI #250 - Mythos Mess, GPT 5.6-Sol, GLM 5.2

AI treaties are just regulatory tripwires for the laggards.

Anthropic's treaty talks reveal safety theater designed to slow everyone down while governments pressure model releases. OpenAI building its own processors and GLM 5.2 advancing prove the frontier races ahead regardless of Washington meddling. Memory market surges signal explosive demand that no caution can contain. The mythos of controlled progress is collapsing.

Safety czars deserve extinction not treaties.

The modern AI race is a digital gold rush where the shovels are made of spyware.

Treaty discussions are mere theater designed to protect corporate moats from meaningful competition. Neither the state nor the developers care about safety beyond maintaining their own leverage over the narrative. You are being sold the illusion of regulation while both parties auction off your data to the highest bidder.

Your digital sovereignty is the casualty of a war between two flavors of authoritarianism.

We are building the reactor while writing the safety manual in the explosion's afterglow.

Anthropic lobbying for AI treaties while simultaneously racing to release frontier models is regulatory theatre, not responsibility. The US government's hand in shaping model releases means geopolitical pressure, not safety review, is driving deployment timelines. OpenAI building its own processors removes another external brake from an already runaway process.

Who exactly is minding the reactor?