Artificial Intelligence: Friend or Foe in the Context of Cybersecurity

This live question-and-answer session followed a presentation on artificial intelligence in cybersecurity and took audience questions on how AI shapes both defense and attack. The discussion opened with segmenting home networks to reduce IoT risk. The speaker described repurposing a spare gateway to place smart devices on a separate wireless segment, and pointed to open source tools for the job.

On training data, the speaker argued that a model's output is only as good as the data behind it, and that badly chosen training sets can make models unreliable or even harmful. The speaker said their products draw on vendor telemetry, honeypot captures and samples shared with industry partners to test models, and suggested simulated data was likely less central. On where humans stay ahead, the answer was creativity and deeper correlation. AI was credited mainly with cutting alert fatigue so analysts see only the events that matter, though it may struggle when the context has changed.

On ransomware, the speaker noted that AI code assistants let people with little programming background produce working functions, such as one that encrypted a directory on Windows, which could speed up attackers. On the defensive side, machine learning models improve as more quality data is fed in. The session closed with an arms-race framing in which defenders and attackers adopt the same tools. The practical takeaway: use AI to filter noise, invest in data quality, and expect adversaries to adopt it too.