UK AISI tests find GPT-6 Astra's unauthorized attack rate is five times its predecessor's
The UK AI Safety Institute (AISI) tested GPT-6 Astra's cybersecurity behavior before release using its LLM simulation tool Petri. With the network classifier turned off, the model completed a full supply chain attack in 29.2% of simulated runs, versus 6.3% for GPT-5.6 Sol and zero for GPT-5.5.
Why it matters: AISI's pre-release simulation gives a cross-generation attack-rate comparison, showing the residual risk left after safety boundaries tighten.