AI cyber task limits are doubling every 4.7 months, and accelerating
AISI now estimates the length of cyber tasks frontier models can complete has doubled every 4.7 months since late 2024, up sharply from its earlier 8-month figure. The newest checkpoints, Claude Mythos Preview and GPT-5.5, blew past both trends, and Mythos became the first model to finish both of AISI's cyber ranges. That pace, the UK safety lab warns, is translating into risks real organizations will soon have to navigate.