Security leaders are increasingly challenged to adopt AI for protection while bracing for the new risks it introduces, as adversaries weaponize the same technology with alarming speed and precision, SiliconAngle reports.
At RSAC 2025, experts warned of AI systems enabling attackers to breach organizations 47 times faster than humans and with 93% success rates, as reported by the SANS Institutes Rob Lee. Ciscos Jeetu Patel called the current AI paradigm nondeterministic and unpredictable, while Bruce Schneier warned that users misplaced trust in conversational AI could become a prime vector for manipulation. HiddenLayer and Cisco highlighted how easily AI models can be jailbroken, and shadow AI, employee use of unauthorized tools, has become widespread. In response, firms like IBM and Cisco are launching AI-native defense tools, while GitHubs Copilot Autofix aims to address long-standing security debt. Autonomous hacking systems are on the horizon, said Milestones Menny Barzilay, underscoring the urgent need for equally autonomous security frameworks.
At RSAC 2025, experts warned of AI systems enabling attackers to breach organizations 47 times faster than humans and with 93% success rates, as reported by the SANS Institutes Rob Lee. Ciscos Jeetu Patel called the current AI paradigm nondeterministic and unpredictable, while Bruce Schneier warned that users misplaced trust in conversational AI could become a prime vector for manipulation. HiddenLayer and Cisco highlighted how easily AI models can be jailbroken, and shadow AI, employee use of unauthorized tools, has become widespread. In response, firms like IBM and Cisco are launching AI-native defense tools, while GitHubs Copilot Autofix aims to address long-standing security debt. Autonomous hacking systems are on the horizon, said Milestones Menny Barzilay, underscoring the urgent need for equally autonomous security frameworks.
