AI’s Hacking Skills Are Approaching an ‘Inflection Point’

AI models are rapidly advancing in their ability to discover cybersecurity vulnerabilities, with some tools now finding previously unknown security flaws that experts missed. RunSybil's AI system Sybil recently identified a complex GraphQL vulnerability requiring deep technical knowledge, while benchmarks show AI models improving from detecting 20% to 30% of known vulnerabilities in months. This creates a double-edged situation where the same AI intelligence used for defensive cybersecurity can potentially be exploited by hackers for malicious purposes.