Back to Geekish
AI Safety - September 17, 2026

Microsoft AI's new code puts containment into the safety fight

Microsoft AI's public safety pitch is not just about making models nicer. It is about keeping increasingly agentic systems controllable, legible, and boxed in.

Illustration of AI policy documents surrounding a contained model core
Image: Geekish-generated illustration based on Microsoft's Humanist AI Code of Conduct and The Verge's Decoder interview.
AI SAFETYGeekish sourced quick read

Original Geekish context based on the sources linked below.

The short version

The Verge published a Decoder interview with Microsoft AI CEO Mustafa Suleyman after Microsoft AI released what it calls a Humanist AI Code of Conduct. The discussion focuses on AI safety, alignment, containment, regulation, and Suleyman's criticism of model-welfare framing.

Containment is the key word

In the interview, Suleyman argues that alignment is only one part of the problem. He says advanced AI systems also need containment: limited agency, controllable behavior, and oversight that humans can actually understand.

The practical standard

One concrete idea discussed is forcing AI systems to communicate in human-readable language rather than opaque machine-to-machine shortcuts. The logic is simple: if humans cannot inspect what agents are saying to each other, meaningful oversight gets much harder.

Geekish take

This is the AI safety debate getting more operational. The old question was "is the model aligned?" The newer question is "can you prove what the agent can do, where it can go, and what it said while getting there?" That is less philosophical, and probably more useful.

Want more tech without boring tech-site energy?

Follow Geekish for sourced quick reads, AI, gadgets, apps, creator tools, and internet culture.

Get the tech drop