Artificial intelligence has in recent years transformed from an emerging technology into one of the world’s largest investment destinations and a potential engine of economic growth. But this rapid growth has come with a worrying reality: regulators around the globe are unable to enact proportionate regulatory frameworks at the same speed. This gap has turned AI regulation from a technical and ethical debate into a complex economic equation.
Stringent laws can raise the cost of developing prototypes, slow product rollout processes, and place a heavy burden on startups. But on the other hand, weak oversight also carries hidden costs that can manifest as large-scale fraud, more sophisticated cyberattacks, automated errors at a million-fold scale, erosion of public trust, and even instability in the financial system.
The risk bill and international warnings
The debate intensified when officials and researchers from AI companies warned about the dangers of more autonomous systems; warnings that coincided with disclosures of incidents involving the use of advanced models in cyberattacks and the increased ability of these systems to perform multi-step tasks with minimal human intervention.
In this context, Volker Türk, the UN High Commissioner for Human Rights, has deemed voluntary self-regulation by companies insufficient and called for mandatory reporting of serious incidents, independent verification of model capabilities, and increased oversight of companies developing the most advanced systems.
Notably, the looming economic risks do not require far-fetched scenarios about superintelligent AI. Ziad Yusuf, a professor and advisor on digital and intelligent systems engineering, told Al Jazeera Net that the first clear risk is cybersecurity and fraud; because AI can significantly increase the capacity to produce malware, forge voices and identities, and execute complex attacks. Banks, insurance companies, ordinary consumers, and critical infrastructure could bear the main costs of these attacks.
Another danger Yusuf points to is “widespread repetition of a single wrong decision.” For example, if a bank uses an uncontrolled model to assess credit or a health institution delegates sensitive decisions to an algorithm, a single error can turn into thousands or even millions of automated incorrect decisions. He also reminds that trust itself has become an economic asset; because broader use of these systems requires their access to sensitive data and performing real tasks, and a decline in public trust can limit technology adoption even amid continuous technical capability progress.
Interlinked risks
Mohammed Al-Suwaisi, a technology specialist, also ranks cybersecurity at the top of economic threats, because economic activity is increasingly moving toward digital systems and AI actors are entering processes that previously required direct human intervention. He says that when these risks are linked to banks, insurers, government agencies, and companies large and small, the sensitivity of the issue multiplies, since any breach or unexpected behavior by a system with broad authority can quickly spread to connected institutions.
Al-Suwaisi does not limit the problem to security and points to another challenge: the limited capacity of many economies and companies to absorb this technology due to a shortage of specialized personnel and the high cost of some models and required infrastructure. He also warns that wide dependence on a small number of providers of language models and cloud computing can create a dangerous form of “technological dependency,” especially when organizations place their sensitive economic and commercial data in systems outside the control of their domestic market.
Another dimension of this equation is the labor market. According to IMF estimates, about 40 percent of jobs worldwide will be affected by AI to varying degrees; either through replacing some tasks or by changing the nature of jobs and the skill requirements.
The cost of safety versus the cost of recklessness
Implementing regulatory controls imposes significant additional costs on companies, including higher-standard prototyping, strengthened cybersecurity, data management, as well as documentation, legal auditing, and continuous post-deployment monitoring. In cases where system risk is higher, independent evaluation may be required prior to release to users.
Al-Suwaisi believes these requirements increase the capital needed before market entry and, due to additional review and approval steps, can delay product time-to-market; a burden that large companies can bear far more easily than startups. But he thinks this cost should be weighed against the potential costs of recklessness; because malfunction, data leakage, legal claims, and even forced withdrawal of a product from the market can be far costlier than early detection of flaws. Yusuf agrees and emphasizes that safety raises development costs in the short term but prevents heavier post-deployment costs in the long term, especially at large scales. According to these two experts, the main solution is “proportionality of rules to risk level”; meaning an application with limited use should not be subject to the same stringent requirements applied to decision-making systems in credit, health, or critical infrastructure.
Investor confidence and global competition
Regulatory environment has lately become a decisive factor in investment and capital allocation decisions. The United States has taken an approach that allows more space for innovation and technical competition; the European Union follows a model classifying systems by risk level and imposes stricter requirements for high-risk systems; and China combines technology regulations with a focus on national security and strengthening domestic capabilities.
Yusuf believes regulation supports investment when it provides clear answers about permissible activities, legal liability, necessary tests, and a risk level that can be priced before injecting capital. But ambiguous, unstable, or disproportionate rules to the real risk level can have the opposite effect and push companies to move development or product launches to other markets. He says, “Capital does not necessarily flee because of the existence of law, but flees from its ambiguity and instability.” Al-Suwaisi also believes legal clarity reduces uncertainty and allows investors to pre-calculate compliance costs, though he warns that complex and costly procedures can become a serious barrier to new entrants.
The risk of entrenching tech giants’ dominance
One of the most worrying potential outcomes of strict regulation is the unintended strengthening of large tech giants’ positions; companies that have the financial capacity to cover compliance costs and sustain large legal and engineering teams, while startups struggle with the same requirements and far more limited resources.
Yusuf warns that imposing a fixed high compliance cost regardless of actual risk can backfire: the regulator’s main goal is to reduce AI risks, but the practical result may be further cementing the dominance of a few large companies. His proposed solution is designing different levels of oversight: baseline requirements in privacy, security, and transparency for low-risk applications, progressively stricter rules in sensitive areas like credit, employment, and health, and ultimately more rigorous testing and monitoring of models capable of conducting cyberattacks or managing critical systems.
Yusuf believes smaller companies can be supported by simplifying documentation requirements, creating regulatory sandbox environments, standard guidelines, and shared or subsidized evaluation facilities to reduce the fixed compliance cost burden on their limited resources. Al-Suwaisi endorses this approach and stresses that simple tools like intelligent translators or administrative assistants should not be treated the same as sensitive systems used in banking, health, and employment. He says regulatory sandboxes can give startups a chance to test products before fully accepting legal responsibilities; because high compliance costs beyond small firms’ capacity can stifle innovation and reinforce the dominance of big players.
The future of jobs and the way forward
The economic consequences of AI regulation are not limited to companies and directly extend to the labor market. The IMF’s estimate that about 40 percent of the world’s jobs will be affected shows public policy faces not only the question of how to prevent unsafe systems, but also how to manage a major economic transition; a transition that will redistribute income, skills, and opportunities across different sectors of the economy and the workforce.
Some analyses warn that poor management of this transformation could turn productivity gains from AI into a factor that widens inequality, especially if the main benefits accrue to owners of capital and technology while the bulk of restructuring costs fall on workers. Conversely, overrestricting the technology can slow potential productivity and innovation gains; therefore, education policies, reskilling programs, and strengthening social safety nets are not substitutes but necessary complements to AI regulation.
Based on this report first published in Al Jazeera, Yusuf believes the success of regulators should not be measured solely by the number of laws enacted or fines imposed, but the real metric is their ability to balance a set of economic and technical indicators; including trends in cyber incidents, frauds and AI-related damages, alongside trends in venture investment, startup formation, technology adoption rates, productivity, market competition levels, compliance costs, and user trust. Continued investment, productivity growth, and the emergence of new businesses coupled with a decline in large costly incidents can signal regulatory success in jointly combining economic growth and technology safety.
Experts emphasize that the emerging experiences in the EU, the United States, and China, as three different regulatory models, will be an important benchmark in coming years to evaluate which approach can better balance innovation, economic competitiveness, and reduction of systemic AI risks.