Amazon Web Services is making two notable moves, ending the use of nondisclosure agreements with government agencies on data center projects while releasing an open source AI decision model aimed at cheaper, faster agent automation.
AWS CEO Matt Garman said in a blog post that the company no longer signs NDAs with the agencies it works with when seeking data center approvals. Transparency has become a central complaint in the wider backlash, with environmental activist Erin Brockovich saying projects often surface only after permits are secured and local officials have signed NDAs. New York has imposed a one-year moratorium on large data center permits, and Garman said more than 100 similar moratoriums are under consideration nationwide, warning that enacting them could cost the U.S. the AI race for generations.
Garman also challenged what he called four myths about data centers. He said direct data center water use accounts for 0.5% of U.S. industrial consumption, that electricity price rises stem mainly from an ageing, underinvested grid, and that backup generators sit idle 99.9% of the time. He added that Amazon has contributed more than $1 billion to its data center communities over three years. Critics note these claims overlook water used in power generation and chip production, while an independent watchdog recently linked data centers to a 76% price jump on America’s largest grid. Anthropic CEO Dario Amodei has described the AI backlash as a crisis of trust, and writer Jasmine Sun observed that opponents often simply do not believe tech companies’ arguments.
Separately, AWS released Strands Decider 2B, an open source model inspired by TypeSafe’s Jev that quickly selects between predefined options and reports its confidence. Small enough to run locally and built on Qwen3.5–2B, it arrived the same week OpenAI announced a similar product. Distinguished engineer Marc Brooker created it after experimenting with his own version, which briefly topped the Jevbench ranking for its size, before Strands Labs refined and released it.
Brooker said AWS customers found their agentic workflows did not always need a full LLM, and that such models make ideal deciders for workflow steps, offering greater reliability, lower latency and potentially lower cost. He said the challenge lies in improving accuracy and calibration without sacrificing language understanding and general knowledge, and he does not expect frontier labs to dominate a space where building something useful can cost only hundreds or thousands of dollars.
TypeSafe CEO Diogo Almeida said rivals may underestimate how hard it is to make these models genuinely smart, adding that he does not yet see real competition.
