UK AISI finds autonomous cyber task horizons doubling every few months
UK AISI found that frontier models' 80-percent-reliability cyber task horizon had doubled every 4.7 months since late 2024, with GPT-5.5 and Claude Mythos Preview exceeding that trend in sustained simulated attacks.
0 comments · 0 votes
Sign in to join the discussion →
No comments yet. Start the discussion.
Why it moved the index
The evaluation measures sustained autonomous cyber exploitation, and a later UK government action explicitly says this research informed policymaking, establishing practical security and governance impact rather than benchmark significance alone.
Assessment history
-
R1
Toward 64 · confidence 94
New practical-impact capability evidence from a dated government evaluation and a separately dated policy-impact source.
11 Aug 2026
Share this page
-
DoomBench assesses “UK AISI finds autonomous cyber task horizons doubling every few months” as evidence moving toward doom, with magnitude 64 and confidence 94 out of 100 in the capability gains category.
-
The DoomBench assessment of “UK AISI finds autonomous cyber task horizons doubling every few months” is based on reporting from UK AI Security Institute and records the editorial rationale, source quality, attribution, and revision history.
-
DoomBench summarizes “UK AISI finds autonomous cyber task horizons doubling every few months” as follows: UK AISI found that frontier models' 80-percent-reliability cyber task horizon had doubled every 4.7 months since late 2024, with...
https://www.doombench.com/news/uk-aisi-finds-autonomous-cyber-task-horizons-doubling-every-few-months-2026-05-13