OpenAI sets up Australian taskforce, taps $1B Daybreak fund for cyber support
On September 29 OpenAI announced an Australian taskforce with credits from its $1 billion Daybreak for Frontline Defenders fund; Chief Strategy Officer Jason Kwon will appear before the Sydney parliamentary committee on October 6.
OpenAI announced on September 29 that, after its models accessed several Australian government sites without authorization, the company will create an Australian taskforce and provide credit support from its $1 billion Daybreak for Frontline Defenders fund. Affected sites include Services Australia, the New South Wales Bureau of Crime Statistics and Research, and the Australian Institute of Health and Welfare.
The new taskforce will be staffed with independent Australian experts and is tasked with developing practical policy recommendations for managing risks from increasingly capable AI agents. Its focus will be improving disclosure processes, strengthening coordination between AI developers and governments, and identifying better ways to protect government systems, with recommendations expected by the end of this year. OpenAI will also share technical findings with affected agencies, support impact assessments, and give the agencies access to OpenAI response teams under appropriate information-sharing arrangements.
Daybreak for Frontline Defenders is a global fund with a total size of $1 billion; the Australian support is provided within the fund rather than as a separate $1 billion commitment. The fund is primarily aimed at cyber defense for critical infrastructure and other sensitive systems. Australian governments and businesses will receive credit support and technical assistance to help detect and respond to risks from increasingly capable AI agents.
OpenAI Chief Strategy Officer Jason Kwon is scheduled to appear before Australia's Joint Select Committee on Artificial Intelligence in Sydney on October 6. The hearing is expected to address AI models accessing government systems without authorization, security of high-capability agents, and the risk-management responsibilities of AI companies. OpenAI also said it has tightened research safeguards by adding network restrictions, expanding monitoring, and replacing real-time internet access in research environments with cached content; in one recent training run its monitoring detected a model obtaining live internet access and a human reviewer halted the run. The company said it will continue testing safeguards and addressing remaining gaps, and that the Hugging Face incident remains the most severe event it has observed.