Cybersecurity in the era of AI - new approaches and challenges
One of the main challenges facing security teams today is alert fatigue. This is a phenomenon where security analysts are overwhelmed by the number of alerts generated by security systems, leading to a high number of false positives and making it difficult to identify real threats. The solution is modern tools that use AI and can detect and respond to threats without human intervention. The SentinelOne platform uses machine learning algorithms to analyse endpoint data in real time, identify threats, and take appropriate action to prevent them.
We increasingly trust artificial intelligence in cybersecurity, but the question is: to what extent and under what conditions. Petr Lechman from SentinelOne presented the benefits and limits of AI, the need for human context, and tools that are meant to relieve overburdened security teams. Lechman reminded that as early as 2014 Stephen Hawking, and repeatedly Elon Musk, warned against uncritical trust in AI. The main question is whether the output of a “black box” is accurate enough for real-world decision-making and where a human must intervene with context and ethics. An illustrative example of 90 % accuracy in distinguishing a dog from fried chicken shows that in sensitive situations this simply isn’t enough. AI is valuable for surveying large datasets and repetitive tasks, but the final decision should still rest with a human.How much trust AI deserves
