Summary
Panelists argued that near-term AI security risk is less “robot overlords” than over-permissioned agents, shadow AI, and old weaknesses accelerated by automation. They urged teams to fix basics like MFA, password managers, pinned GitHub Actions, and agent permissions before productivity-driven deployments expand attack surfaces.
Key Topics
5 key topics from this session at HumanX Amsterdam 2026. Thicker branches were mentioned more often in the talk.
Reality Check
Click a topic to open it
Key Takeaways
- Enable MFA before worrying about advanced AI attack scenarios.
- Limit agent permissions and contain blast radius by default.
- Inventory shadow AI before contractors connect it to source code.
- Pin GitHub Actions and fix boring security basics first.
- Treat AI deployments as attack-surface expansion unless controlled.
“I think we should assume at this point that everything hackable can and will be hacked”
Share
Up Next
Y
WatchNext in agenda
The Agentic Cloud: From Intelligence to Action
YYe Huang
agentic aialibaba cloud
LN
WatchAlso on agentic ai
How the World's Largest Food & Beverage Company Accelerates Digital Innovation
Luca Dell’Orletta, Niall Firth
agentic aienterprise ai
JLT
WatchAlso on agentic ai
The Next Generation of Digital Experiences
John Imah, Linda Tong, Tom Allen
aidigital experiences