AI: Who's watching whom?

The session starts by treating generative AI as a fast-spreading technology, comparing how quickly consumers adopted recent AI tools with earlier shifts such as the internet and smartphones. It argues that employees are already experimenting with these tools at work, often outside any corporate oversight, so the question for organizations is how to use AI safely rather than whether it will be used at all.

The middle section covers the threat side. Phishing lures that once gave themselves away through poor spelling can now be polished and personalized, and voice cloning and deepfakes make help desk and executive impersonation more believable. The talk also points to attackers using AI to act on newly disclosed vulnerabilities in internet-facing devices sooner than patching cycles allow, to info-stealer malware harvesting tokens and passwords, and to sprawling cloud and SaaS environments where many tools do not share data. Staffing shortages, change freezes and budget limits are presented as reasons defenders fall behind. Employees pasting sensitive material into public AI services is highlighted as a data leakage risk that many organizations cannot see.

The closing portion looks at AI on the defensive side: assistants built into security products that help analyze logs, tune policies, clean up unused rules and generate audit reports. It describes a path from human-approved suggestions toward more automated responses. For practitioners, the takeaways are to improve visibility across tools, train users on safe use of generative AI, add input controls before building internal models, and use AI to ease limited staffing.