'This version of GPT-5.4 lowers the refusal boundary for 'legitimate' cybersecurity work', OpenAI said.
Read more: After Anthropic, OpenAI launches cyber-specific AI model
April 15, 2026 - BY Silicon Republic