CAISI Works with OpenAI and Anthropic to Promote Secure AI Innovation

NIST – AI News (topic 2753736)(US) 25 Sep 2025 62

US-UK joint AI security evaluations of frontier models set a practical benchmark—Australian agencies and AISI should note what this partnership is producing.

  • CAISI worked with OpenAI and Anthropic to identify security vulnerabilities and improve AI security measurement.
  • Evaluations were completed in partnership with the UK AI Security Institute, signalling ongoing Five Eyes-adjacent AI safety cooperation.
  • Australia's AISI is not mentioned; this bilateral US-UK arrangement may inform where Australia sits in frontier AI security collaboration.
  • Monitor Australia's AISI and DISR policy teams may want to monitor what security findings and best practices emerge from CAISI-UK AISI evaluations, given their likely influence on international AI security standards.
  • Consider Agencies could consider whether Australia's bilateral AI safety arrangements with the US and UK align with, or risk lagging behind, the depth of US-UK joint evaluation work now underway.

Implications are AI-generated. Starting points, not advice — see methodology for how they're framed.

View original source