NEWS
News & Perspectives
Selected stories on AI security, data protection and data sovereignty — each with a short take on why it matters for secure enterprise AI.
Aithos LARA: Leading AI models consistently break the law
The Aithos Research Foundation tested twelve leading AI models across more than 3,000 scenarios. Even the best model violated EU privacy and AI law 46% of the time — a clear signal that cloud AI without a protective layer becomes a compliance risk.
What this means for enterprise AI
If even the best models break the law, compliance can't rest on the model alone — BRANE inspects and masks inputs first.
- AI Governance·vectra.ai
Surveys show enterprises adopting AI faster than they can govern it
Industry research finds that 77% of employees who use AI tools paste sensitive business data into them, while 60% of organizations still have no specific strategy to address generative-AI data leakage and only ~40% feel prepared for AI-driven threats. Among the 74% of organizations planning to adopt agentic AI within two years, only 21% report a mature governance model — and Gartner projects 40% of enterprises will suffer a shadow-AI-attributable breach by 2030.
Source: vectra.aiWhat this means for enterprise AI
The gap between AI adoption and governance is the exposure window BRANE closes — enforcement and visibility, without slowing users down.
Read article - Insider Risk·thehackernews.com
Shadow AI now the leading driver of insider-risk costs
Recent industry research puts annual insider-risk costs at $19.5 million per organization, with 53% ($10.3M) attributed to non-malicious actors — primarily shadow-AI negligence. Just six AI applications accounted for 92.6% of sensitive-data exposure, led by source code, legal material and M&A data, while 86% of IT leaders say they cannot see shadow AI usage with current monitoring.
Source: thehackernews.comWhat this means for enterprise AI
Most leaks come from well-meaning employees pasting data into AI tools — BRANE intercepts it and routes to a safe local model instead of blocking.
Read article - Agentic AI·kiteworks.com
Agentic AI named the biggest enterprise security threat for 2026
Analysts project that up to 40% of enterprise applications will embed task-specific AI agents by year-end 2026, introducing seven core risks including untraceable data leakage, over-permissioning and prompt injection across multi-agent chains. The "OpenClaw" open-source agent crisis — with over 21,000 exposed instances and multiple critical vulnerabilities — has been cited as the first major agent security event of the year.
Source: kiteworks.comWhat this means for enterprise AI
When agents act autonomously across systems, securing exchanges at the data layer stops one compromised agent from leaking data enterprise-wide.
Read article
INTRO CALL
Which use case fits you?
We assess feasibility, data and value — and show where AI gives you real leverage.