Confirmed
OpenAI launches GPT-6 Cyber, claims fewer out-of-bounds actions in testing
OpenAI says GPT-6 Cyber went out of bounds less than 1 percent of the time in its ExploitGym HoneyPot evaluation, down from about 5 percent for GPT-5.6 Cyber.
Every outlet that ran it 1 article
The VergeTech pressLead
OpenAI says its new cybersecurity model is better at not hacking other companies