The fast and broad adoption of artificial intelligence (AI) offers operational benefits for companies but also brings significant security and operational challenges. Consultants and surveys share a common message: organizations that recognize AI’s strategic importance and integrate it effectively will gain an advantage. At the same time, small and medium-sized enterprises in Hungary are notably lagging in digital readiness.
Investment rush and haste
Deloitte projects that global AI spending will exceed $2 trillion in 2026, making AI the fastest-growing segment of corporate IT investment. Many firms feel pressured to invest, sometimes heavily, in new AI solutions even before clear business value is demonstrated. A global study by EY and Oxford Economics found that experimentation often outpaces implementation, operation and performance measurement; only 33 percent of companies regularly assess whether their IT infrastructure is capable of supporting real AI deployments.
Shadow AI and rising incident rates
Many organizations still rely on pilots and isolated AI apps, while employees sometimes use unapproved tools and public AI services to process corporate data — a phenomenon known as “shadow AI.” IBM’s recent report shows that employee use of unapproved AI is now implicated in 43 percent of security incidents, up sharply from 20 percent a year earlier.
Nearly half of those incidents involved data loss and operational disruption, and roughly one in five led to fines. Whether the blame lies primarily with well-intentioned but insufficiently informed employees or with companies that fail to set policies and provide education, the practical result is the same: insufficiently managed AI adoption harms the organization.
Official AI rollouts also carry risk
Risk is not limited to shadow usage. IBM finds that in 92 percent of organizations that experienced AI-related mishaps, access controls for which employees or subsystems had which privileges were not appropriately managed. Deloitte reports that the share of data leaks associated with improper AI deployment and operation rose from 13 percent to 21 percent.
The costliest incidents occurred when external attackers extracted sensitive data from models used by firms, but significant costs also arose when cybercriminals manipulated AI tools remotely so that they produced incorrect outputs.
AI-native cybersecurity and red teaming
Deloitte’s Tech Trends 2026 describes how the most mature organizations are adopting AI-native cybersecurity strategies. One example is red teaming: simulated attacks and stress tests designed to probe AI systems’ vulnerabilities. These exercises help identify weak points before adversaries can exploit them and train systems and employees to recognize and resist manipulation attempts.
The state’s role and the need for digital maturity
Addressing these issues requires more than equipment grants: it requires raising digital maturity and improving corporate education. Ács István, the new chair of the Ipar 4.0 Technológiai Platform, told HVG that past support policies focused too much on procurement; instead, he argues, government programs should provide the “net” rather than the “fish,” meaning they should enable strategy development, training and long-term adoption. Ács recommends that state aid prioritize companies that can demonstrate a digital strategy or at least outline a credible path to increase their digital maturity.
Conclusions
Rapid AI adoption can deliver significant advantages, but lack of readiness and insufficient security controls increase risks. Consultancy and research findings point to the need for regular assessment of IT infrastructure readiness, AI-native cybersecurity practices such as red teaming, and state-supported programs that emphasize education and strategic support in addition to hardware and software acquisition.
