# amd

OpenAI's planned loosening of adult interaction restrictions alongside a 6‑gigawatt GPU pact with AMD signals accelerating capability and expanding ethical obligations. Simulations with 44,000 AI agents and new evidence of easy model poisoning underscore why governance, grid access, and resilience now rival model quality.

Reports of a 25% AMD surge tied to an OpenAI stake and multi‑gigawatt GPU plans collided with Michigan officials drafting data center guidance, exposing water and power bottlenecks. Practitioners challenged claims that models can detect evaluations and highlighted chaos in workplace deployments, pushing governance and measurement to the foreground. Meanwhile, creators tested image‑protection techniques as open models topped text‑to‑video benchmarks, underscoring tensions between openness and control.

Escalating compute ambitions are colliding with real-world infrastructure, with a Fortune investigation projecting a 17GW energy appetite for advanced models while AMD's GAIA broadens access through new Linux support. Governance tensions are sharpening, as a preliminarily approved $1.5B Bartz v. Anthropic settlement and looming privacy policy changes put data rights and accountability at the forefront. Emerging capabilities from synthetic companions to AI-designed bacteriophages underscore the urgency of safety and incentive design.