The guardrails are not keeping up with the usage
AI incidents hit a record in 2025, hallucination rates run as high as 94%, and one in nine organisations still has no responsible-AI policy. This is the security gap, measured.
Adoption is racing ahead; oversight is walking. The AI Index records a record year for documented AI incidents, wide and unpredictable error rates, and a transparency picture that is getting worse, not better. The space between how much AI is used and how well it is governed is exactly where data leaks, breaches and compliance failures happen.
362
documented AI incidents in 2025
up from 233 in 2024, a 55% rise
435
AI incidents in the peak month
January 2026
22–94%
hallucination rate across 26 leading models
error is the rule, not the exception
11%
of businesses still have no responsible-AI policy
down from 24%, but still roughly one in nine
17%
growth in dedicated AI-governance roles in 2025
Falling
model transparency scores
knowledge gaps and regulatory uncertainty are the top barriers
Source: Stanford HAI 2026 AI Index
Responsible AI chapterWhat it means for you
This gap is the whole reason Pryvan exists.
Incidents rising 55% in a year, error rates up to 94%, and one in nine companies with no policy at all: that is what an unmanaged AI rollout looks like. The fix is not to slow down, it is to run AI where you can see it. EU-hosted open models on infrastructure you control turn an invisible risk into a governed one, with audit logs and a compliance pack to prove it.
Bring AI into your business, without giving up your data.
Join the waitlist. We're onboarding GDPR-sensitive SMEs across Europe.