OpenAI Launches GPT-6 Astra, Scores Perfectly on Cybersecurity Exploit Benchmark

OpenAI has introduced GPT-6 Astra, an advanced AI model described as the "world's most intelligent and aligned." Astra achieved a perfect 100% score on ExploitBench, a benchmark assessing an AI's ability to convert software vulnerabilities into working exploits, significantly surpassing its predecessor. OpenAI also confirmed Astra meets the "Critical" cybersecurity capability threshold under its Preparedness Framework, raising questions about AI's dual-use potential in security.