On 2 August 2026, another important stage in the implementation of the European Union’s AI Act began. This does not mean, however, that all requirements concerning high-risk systems became applicable on that date. Among other provisions, rules on the transparency of certain AI systems and AI-generated content began to apply, while the EU AI Office and the relevant national authorities received broader supervisory and enforcement powers. The full requirements for some categories of high-risk systems will apply at later dates.
Investors, however, began factoring regulatory risk into their decisions much earlier. A study conducted by the appliedAI Institute among 113 European startups developing AI solutions and 15 venture capital funds found that, according to respondents and the report’s authors, between 33% and 50% of the systems analysed could potentially be classified as high-risk. This was significantly higher than the 5% to 15% assumed in earlier regulatory impact assessments.
These findings should nevertheless be interpreted with caution. They do not reflect classifications made by supervisory authorities or demonstrate an actual outflow of capital following the implementation of the regulation. Instead, they primarily reflect the expectations of startups and investors regarding the new legal requirements. The study covered only part of the European market, while respondents’ understanding of the details of the AI Act may also have varied.
Even so, the study points to a clear trend: venture capital funds expected startups developing high-risk systems to become less attractive investment targets. Most of the investors surveyed anticipated a decline in both the likelihood of investing in such companies and their valuations. Assessments of low-risk solutions were considerably more stable, with some funds even expecting their investment appeal to increase slightly.
This is not merely a question of formal compliance. During due diligence, investors assess not only a product and its commercial potential, but also the cost of regulatory compliance, the risk of delays in bringing it to market, the company’s ability to scale and its potential legal liability. For the founder of an AI startup, determining whether a solution falls under the rules governing high-risk systems may therefore have a direct impact on the company’s valuation and financing terms.
High-risk systems may include solutions used in areas such as recruitment and employee management, education, access to essential services, creditworthiness assessments, biometrics, critical infrastructure, law enforcement, migration and the administration of justice.
The use of AI to profile customer behaviour or increase sales does not, however, automatically result in a system being classified as high-risk. Each case must be assessed in light of the system’s intended purpose, the way it is used and its impact on the rights and circumstances of individuals.
For providers of high-risk systems, the regulation may introduce additional obligations relating to risk management, data quality, technical documentation, record-keeping, human oversight, cybersecurity, conformity assessments and post-market monitoring. The scope of responsibility will nevertheless depend on the role of the organisation concerned. Different obligations may apply to the system’s provider and to a company that merely deploys or uses it.
The way penalties are presented also requires clarification. The maximum fine of EUR 35 million or 7% of global annual turnover does not apply to every infringement of the AI Act. The highest penalties are intended primarily for the use of prohibited AI practices. Lower penalty thresholds may apply to failures to comply with other obligations. In the case of small and medium-sized enterprises, the amount of the fine should also take account of the principle of proportionality.
The scale of the potential burden helps explain why some startups participating in the study said they were considering relocating their operations outside the European Union or discontinuing the development of AI solutions. These were, however, declarations concerning possible future decisions rather than confirmed cases of relocation or cancelled investment.
On the other hand, the AI Act may strengthen the position of companies that establish risk-management systems and prepare the necessary documentation at an early stage. Demonstrated compliance may reduce uncertainty during due diligence, facilitate negotiations with large corporate clients and improve credibility in regulated industries and public procurement procedures.
This does not mean that compliance automatically guarantees access to financing. It is instead one of several factors considered by investors alongside the quality of the technology, the capabilities of the team, the business model and the company’s ability to scale.
The AI Act should therefore be viewed more broadly than simply as a source of bureaucracy. For some startups, the regulation will increase costs and extend the time required to bring a product to market. For others, it may become a barrier to entry that protects them from less-prepared competitors.
From an investment perspective, the key consideration remains the expected return in relation to regulatory, technological and business risk. Ultimately, this calculation will determine company valuations and influence how capital is allocated.
Source: CEO.com.pl
