A recent report by Trustmarque reveals that many organizations are failing to address AI-specific risks during the development and deployment of artificial intelligence systems.

Despite the unique challenges AI introduces—such as model bias, lack of transparency, and hallucinations—business leaders continue to rely on traditional software development frameworks, which are ill-suited for managing these emerging threats.

The report highlights that while 93% of organizations are already using AI in some form, very few have adapted their existing processes to handle AI-related vulnerabilities. Shockingly, only 7% have fully implemented governance structures, and over half operate with either no governance or extremely limited oversight.

Moreover, just 8% have successfully integrated AI governance into their software development lifecycle (SDLC), with most describing their efforts as fragmented or inconsistent across departments.

AI models carry inherent dangers, including biased outputs and hallucinations—fabricated responses presented as facts. Although many companies perform security checks or monitor for anomalies, only 28% actively test for bias during development, and a mere 22% assess model interpretability.

One major obstacle is inadequate infrastructure. Only 4% of organizations say their data and technical environments are fully equipped to support large-scale AI initiatives. Crucially, essential components like model registries, audit trails, and version control are often handled manually or not at all.

“AI adoption is clearly outpacing governance,” said Seb Burrell, head of AI at Trustmarque. “With 93% of organizations using AI but only 7% having comprehensive governance in place, there’s a significant disconnect. Innovation is moving faster than systems and support structures can keep up.”

He added, “Development teams lack the necessary tools and infrastructure, and this is worsened by insufficient leadership support for building strong governance frameworks.”

Accountability remains diffuse, with only 9% of respondents indicating alignment between IT leadership and governance responsibilities. A full 19% reported having no designated owner for AI governance.

Most governance initiatives are led at the departmental level rather than driven by enterprise-wide strategy. Continuous monitoring is rare—only 18% of organizations use KPIs to track AI performance and compliance over time.

Cross-functional collaboration is also lacking. Legal, ethics, and HR teams are only occasionally consulted in AI decision-making (40%), and just 20% have established formal, multi-departmental governance bodies.

According to the report, this disconnect between ambition and implementation represents the biggest hurdle for enterprises leveraging AI today. The authors recommend aligning AI initiatives with core business goals, embedding governance directly into development workflows, and investing in both technological infrastructure and workforce capabilities.

They also emphasize the need for shared accountability across functions, rather than siloed responsibility.

“Organizations are rolling out generative AI faster than they can manage it, and the risks are tangible,” warned Burrell. “Without proper oversight, we’re seeing increased exposure to privacy violations, operational failures, ethical lapses, and erosion of stakeholder trust. These aren’t just regulatory concerns—they’re fundamental threats to long-term business stability and credibility.”

The above is the detailed content of Organizations face ticking timebomb over AI governance. For more information, please follow other related articles on the PHP Chinese website!