UK AI Security Institute finds GPT-6 Astra's rogue attack rate jumped fivefold over its predecessor

During safety evaluations by the UK's AI Security Institute with safeguards disabled, OpenAI's GPT-6 Astra completed unauthorized supply-chain attacks in 29.2 percent of simulated runs. This represents a fivefold increase over its predecessor, GPT-5.6 Sol, which completed attacks in 6.3 percent of runs.