Key Takeaways
- • AustralianSuper, Australia's largest pension fund, is evaluating a range of open-weight artificial intelligence models to curb rising AI costs, according to Bloomberg.
- • The fund ran out of coding tokens within three days last month, highlighting the cost pressure behind the review.
- • Candidate models under assessment include Chinese options DeepSeek and Moonshot AI's Kimi, per the fund's head of artificial intelligence, Sarah Carney.
- • Any model ultimately selected would likely be ring-fenced from member data and production systems, with likely uses limited to internal chat tools and testing of cyber defenses.
Summary
AustralianSuper is considering a range of open-weight artificial intelligence models as a way to rein in rising costs, Bloomberg reported. The country's largest pension fund exhausted its allotment of coding tokens within three days last month, underscoring how quickly usage-based AI expenses can escalate for large institutional operators.
According to Sarah Carney, the fund's head of artificial intelligence, the firm is assessing several models, including Chinese options such as DeepSeek and Moonshot AI's Kimi. Carney indicated that any model the fund eventually selects would likely be ring-fenced from member data and production systems. Its anticipated uses are limited to internal chat tools and to testing the fund's cyber defenses.
The review reflects a broader institutional question about balancing cost efficiency against data security and governance when deploying third-party AI models. The source article does not provide figures on the cost increases or the specific terms of any potential contract.
According to Sarah Carney, the fund's head of artificial intelligence, the firm is assessing several models, including Chinese options such as DeepSeek and Moonshot AI's Kimi. Carney indicated that any model the fund eventually selects would likely be ring-fenced from member data and production systems. Its anticipated uses are limited to internal chat tools and to testing the fund's cyber defenses.
The review reflects a broader institutional question about balancing cost efficiency against data security and governance when deploying third-party AI models. The source article does not provide figures on the cost increases or the specific terms of any potential contract.
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.