AI Security from Cisco

I just watched a video on Simply Cyber with Gerald Auger, who interviewed Jeetu Patel, the Senior Vice President for Product at Cisco. They talked about the security issues when adopting AI and what users can do to shore up their systems. The three biggest challenges that security departments are finding are lack of experienced employees, insufficient number of employees to diagnose the number of attacks they are receiving, and too many alerts to parse through for assignment. The most challenging difference that makes security so difficult involves the network architecture. Into an architecture of 3 or 4 layers of deterministic software, we have introduced one or more models which are inherently nondeterministic. This means that the output is always different, so security software must be rethought from the beginning. Another change on the near horizon is agentic software which translates to many domain-specific agents taking charge of jobs that need to be done under the control of an orchestrator agent. This network has both the domain agents working with the security agents to complete the job in a safe manner. If you want to hear much more, check out this video.