The U.S. government is launching a new coordination framework called GoldEagle to improve detection of critical infrastructure vulnerabilities, set priorities and accelerate patching. The initiative uses existing federal authorities and resources to connect open-source software managers and critical infrastructure operators through a single reporting and response pipeline. The White House said it has begun receiving and prioritising reports, but gave no details on participating companies, AI models or prioritisation methods.

Concern is growing over cyber threats that abuse advanced AI models, prompting governments to speed up policy responses. South Korea’s Ministry of Science and ICT and KISA issued manuals, guidelines and an AI security red-teaming guide. China warned of a backdoor risk in Anthropic’s Claude Code, while Alibaba banned its use. U.S. agencies are reported to use Anthropic models for code scanning and cyber operations support.