A recent survey by Wolters Kluwer Health reveals that over 40% of healthcare workers are aware of colleagues using unauthorized artificial intelligence tools, known as “shadow AI,” posing potential risks to patient safety and data privacy. Dr. Peter Bonis, chief medical officer at Wolters Kluwer, emphasized that while these tools may be beneficial for individual users, their safety and efficacy have not been assessed by healthcare organizations, raising concerns about misleading or inaccurate information that could harm patients. The survey highlights that many healthcare professionals are not fully aware of their organization’s AI policies, with only 29% of providers familiar with the main guidelines.
A recent study reveals that Google’s AI Overviews cite YouTube more frequently than any medical website when addressing health-related queries. Researchers at SE Ranking analyzed over 50,000 health searches and found that YouTube accounted for 4.43% of all citations by AI Overviews, significantly outpacing traditional medical sources like hospitals or government health sites. The findings raise concerns about the reliability of the information being provided, as YouTube is a general-purpose platform where anyone can upload content, including unverified sources. The dominance of YouTube as a reference suggests a troubling trend where popularity may outweigh medical authority in delivering health information to the public.
A recent report from Harvard Business Review Analytic Services reveals that while enthusiasm for agentic artificial intelligence is growing, many organizations lack the readiness to implement it effectively. Despite high expectations for productivity gains, only a minority of companies are using agentic AI at scale, with significant gaps in governance and workforce skills hindering progress. The report indicates that organizations must redefine success, invest in personnel, and establish strong governance to bridge this gap. This framework sets clear limits on AI autonomy and emphasizes the importance of formalizing AI agents as digital identities to mitigate security risks, according to industry leaders and reports from Accenture and Okta. Their research highlights that while over 90% of organizations are using AI agents, very few have established effective oversight, suggesting that clearer rules are essential as these technologies increasingly influence decision-making with real-world consequences.
Why do we care?
AI already shaping decisions in healthcare, public information, and enterprise operations—often without formal oversight.
Clinicians are using unsanctioned AI tools because they help them work faster, even when policies are unclear. Google’s AI summaries increasingly rely on general platforms like YouTube, signaling a shift in how authority is inferred by models. Meanwhile, organizations are deploying AI agents with limited rules about what those agents are allowed to do.
The common issue isn’t adoption—it’s accountability, as AI systems are influencing outcomes before organizations define who owns the result. In practice, this shows up when a client acts on an AI-generated recommendation, a model hallucination informs care or compliance, or an AI agent triggers a workflow—and the MSP is asked who approved it, validated it, or insured it.
For IT service providers, this represents a growing risk. Clients will expect guidance, control, and liability clarity around AI-driven decisions. Those who can provide governance—not just tools—will create real value.