Key Takeaways
- • AustralianSuper, Australia's largest pension fund, is evaluating a range of open weight AI models, including Chinese options such as DeepSeek and Moonshot AI's Kimi, to curb rising AI costs.
- • The fund reportedly ran out of coding tokens within three days last month as AI usage costs spiked, according to Bloomberg.
- • Sarah Carney, the fund's head of artificial intelligence, said any model selected would likely be ring-fenced from member data and production systems.
- • Likely internal use cases include chat tools for staff and testing the fund's cyber defenses.
Summary
AustralianSuper is considering open weight artificial intelligence models as a way to rein in rising costs, according to Bloomberg. The country's largest pension fund recently exhausted its allocation of coding tokens within three days last month, highlighting how quickly AI consumption costs can escalate for large institutional users.
According to Sarah Carney, the fund's head of artificial intelligence, the firm is assessing a range of models, including Chinese options such as DeepSeek and Moonshot AI's Kimi. Carney said any model the fund ultimately selects would likely be ring-fenced from member data and production systems, limiting its exposure to sensitive information.
The prospective use cases are narrow. Carney said the models would likely be deployed for internal chat tools and to test the fund's cyber defenses. The consideration of Chinese-developed models by a major Australian superannuation fund is notable given the sensitivity around data security and geopolitics in institutional technology procurement, though the primary source does not indicate a final decision has been made.
According to Sarah Carney, the fund's head of artificial intelligence, the firm is assessing a range of models, including Chinese options such as DeepSeek and Moonshot AI's Kimi. Carney said any model the fund ultimately selects would likely be ring-fenced from member data and production systems, limiting its exposure to sensitive information.
The prospective use cases are narrow. Carney said the models would likely be deployed for internal chat tools and to test the fund's cyber defenses. The consideration of Chinese-developed models by a major Australian superannuation fund is notable given the sensitivity around data security and geopolitics in institutional technology procurement, though the primary source does not indicate a final decision has been made.
Go deeper than the headline
Research every firm in this story, track the ones that matter and get the daily briefing — free for 7 days.