AI Navigate
アップデート最新記事AI大全カオスマップ

US government now has pre-release access to AI models from five major labs for national security testing

THE DECODER / 5/6/2026

📰 NewsDeveloper Stack & InfrastructureSignals & Early TrendsIndustry & Market Moves
Read original →
共有:

Key Points

  • The U.S. Department of Commerce is expanding AI safety testing by adding pre-release access to AI models from five major labs for national security use.
  • Anthropic and OpenAI were followed by agreements with Google DeepMind, Microsoft, and xAI through the Center for AI Standards and Innovation.
  • The provided models include reduced safety “guardrails” so they can be tested in classified environments.
  • The move is driven by rising cybersecurity threats and an increasingly competitive technology race with China.
  • The agreements reflect a broader effort to operationalize AI safety evaluation under government oversight before models are broadly deployed.

The US Department of Commerce is expanding its AI safety testing: Following Anthropic and OpenAI, Google Deepmind, Microsoft, and xAI have now signed agreements with the Center for AI Standards and Innovation. The companies provide models with reduced safety guardrails for testing in classified environments amid growing cybersecurity risks and an intensifying tech race with China.

The article US government now has pre-release access to AI models from five major labs for national security testing appeared first on The Decoder.

関連おすすめサービス

※当サイトはアフィリエイト広告を利用しています
🎧

Notta搭載AI議事録イヤホン ZENCHORD1

AI時代の仕事術。Notta搭載で会議の議事録を自動生成するスマートイヤホン。

🎙️

AI搭載ボイスレコーダー Plaud

世界100万人が愛用。AIで文字起こし・要約を自動化するボイスレコーダー。

🖼️

画像高画質化AIツール Aiarty Image Enhancer

AIで画像を高画質化。写真・イラストを簡単にアップスケール。