OpenAI launches GPT-5.6-Cyber with reduced refusals, 95% completion on advanced cybersecurity tasks

OpenAI launches GPT-5.6-Cyber with reduced refusals, 95% completion on advanced cybersecurity tasks

OpenAI unveiled GPT-5.6-Cyber achieving 95% on exploit-development benchmarks, available only to Daybreak Red-approved organizations. Security leaders must now evaluate access tiers and budget dedicated AI capabilities as OpenAI pivots toward domain-specialized models.