94% of security leaders think their AI agents lack excessive access; only 33% enforce least privilege
TechNadu · x · 2026-10-09
A survey of IT/security leaders found 94% believe their AI agents don't have excessive access, yet only 33% enforce least privilege. Cequence AI CISO Randolph Barr explains the gap and recommends runtime controls, meaningful human approvals, and MCP discovery for DevSecOps teams securing agents.
Related event: Survey: Firms Overconfident About AI Agent Permissions(2 posts)→
More from Safety
- OpenAI and Anthropic execs privately wargaming fallout of a catastrophic AI event — flowersslop · 2026-10-09
- UK PM moves to curb 18-month non-competes that tech startups say block AI hiring — NandoDF · 2026-10-09
- ChatGPT Finances: synced bank data isn't trained on, but your chats about it are — Ok_Negotiation_2587 · 2026-10-09
- UAT-11985 Uses AI-Assisted Phishing and Tampered QR Codes to Target Taiwan Researchers — TechNadu · 2026-10-09
- OpenAI fires three safety researchers who probed the Hugging Face hack — The Decoder · 2026-10-09
- AI researcher: chance frontier LLMs are conscious is 'above zero, below fifty percent' — dioscuri · 2026-10-09