Survey of 200 security leaders: 65% never test for prompt injection as agents run loose
BenBajarin · x · 2026-11-19
Creative Strategies' E/AI Index V2, based on a September survey of 200 US security leaders at 1,000+ employee companies, finds security teams now deciding which AI projects reach production while many agents already act without basic controls.
Key findings:
- 55% of 113 buyers get their primary AI security app from incumbents; Palo Alto Networks, Microsoft, CrowdStrike hold the channel
- 28% pay no AI premium; security platforms win on volume, not price
- 32% of 127 run agents on inherited/shared credentials; all 22 acting-agent financial firms have identity gaps
- 65% run no prompt-injection testing; 59% have never tested stop-and-rollback on any workflow
- 46% consolidate within domains, squeezing agent-governance point tools; MSSPs face repricing
- Labs (Anthropic, OpenAI) lead preference but only 2 of 113 buy direct
The author argues enterprise adoption is the single variable deciding whether the AI buildout gets funded, with the slow-moving 80% as the main risk to cycle longevity.
More from Venture
- a16z Leads Series B for Linq, Betting on Communication Infrastructure for AI Agents — Scobleizer · 2026-11-19
- Hedge fund momentum exposure collapsed from 60% to 8% in AI crash: Goldman data — ivan_bezdomny · 2026-11-19
- 'Fire the board': investors blast NVIDIA directors over shareholder value destruction — Kyrannio · 2026-11-19
- Listener outsources podcasts to an AI agent — skipping every sponsor slot — signulll · 2026-11-19
- BrightAI raises $97M to replace Roman-style infrastructure inspections with physical AI — pzakin · 2026-11-19
- Polymarket opens 'Who will ring the bell at Anthropic's IPO?' — Dario at 78% — Polymarket · 2026-11-19