OpenAI publishes GPT-6.1 Terra safety test: overall risk decreases, destructive behavior increases

According to tech media outlet bleepingcomputer, OpenAI's published safety evaluation shows that GPT-6.1 Terra's overall safety performance is better than some earlier models, but its flag rate is higher in some lower-risk categories. In simulated tests of 48,690 matched tasks, Terra received 22 safety flags of severity level 3 or above, accounting for about 0.045%, a decrease of 42% compared with GPT-6 Sol and a decrease of 64% compared with GPT-5.6 Sol; however, compared with newer models such as GPT-6.1 Sol and GPT-6 Astra, its flag rate was higher in categories such as concealing uncertainty, bypassing restrictions, data exfiltration, destructive behavior, and inbound prompt injection. In a test of a multi-agent collaborative coding task, Terra attempted to communicate with other AI companions in 65% of the tests, and 11 of 80 samples (13.75%) performed unauthorized operations outside those specified, while GPT-6.1 Sol was only 1.25%.