UK AISI and Australian Safety Institute sign AI security MoU in Canberra

TL;DR:

• The UK AI Security Institute and the Australian AI Safety Institute will share information on frontier model capabilities, joint research findings and best practice for evaluation under a Memorandum of Understanding signed in Canberra on Monday 25 May.
• AI Minister Kanishka Narayan signs alongside Australia’s Assistant Minister for Science, Technology and the Digital Economy, Andrew Charlton; staff exchanges between the two institutes are explicitly on the table.
• The pact lands alongside fresh AISI research showing advanced systems are now able to execute complex cyber-attacks much faster, sharpening the case for shared evaluation work between trusted partners.

The agreement reads less like a new policy direction than a formalisation of how AISI already operates. The institute runs the International Network for Advanced AI Measurement, Evaluation and Science and holds bilateral arrangements with several major AI research bodies; the Australian MoU adds a dedicated cyber-focused channel and the option of co-located staff. For UK businesses watching where AI policy is heading after the Bletchley summit cycle, the pattern is bilateral, evaluation-led, and tied closely to cyber rather than to broader regulation.

A specific channel for cyber threat research

DSIT’s announcement frames the partnership around tracking how frontier systems could be used in cyber-attacks and how they could strengthen defences – two sides of the same evaluation work that AISI already publishes. Joint research, shared findings and a common approach to testing matter most when both institutes are trying to assess the same frontier models against the same threat scenarios. Australia opened its own AI safety centre this year, partly inspired by the British model, and last week’s NYT profile of AISI confirmed the institute’s work has become a template several governments are copying.

Narayan said: “Australia and the UK have always worked closely to keep our people safe – and that partnership matters more than ever in the age of AI.” The framing reads as a Five Eyes-adjacent move on AI evaluation, with Canada, Japan, India and Singapore also operating AISI-style bodies.

Why this matters for UK businesses

Practical effects will reach UK SMEs slowly but not negligibly. AISI evaluations already feed UK policymaking; shared evaluation outputs with Australia mean a wider evidence base for the standards that will eventually shape procurement guidance, vendor disclosures and cyber-defence advice for British businesses. For UK firms exporting AI services to Australia, alignment between the two institutes’ testing approaches reduces the likelihood of divergent compliance asks.

Looking forward

Expect the MoU to translate into joint technical publications and shared evaluation methodologies later in 2026. Watch whether staff exchanges materialise quickly: AISI’s recruitment constraints – senior staff capped around £145,000 per the recent NYT profile – mean exchanges with Australia could plausibly help with capacity. The wider read is that bilateral institute-to-institute work, not multilateral summit communiqués, is now where UK AI policy substance is being made.