OECD: 61 percent of venture capital went to AI. Investors still cannot compare companies’ safety

An analysis on the relationship between investment and trustworthy artificial intelligence was published on the OECD.AI platform. The article draws on a June meeting of an expert group from the international GPAI partnership and examines how investors can take the safety and accountability of technology companies into account in their decisions. According to data cited in the analysis, about 61 percent of global venture capital went to AI-focused companies in 2025. Such concentration creates exceptional pressure for rapid growth and technology expansion. At the same time, investors lack sufficient comparable information about how individual companies manage risk. Many firms publish general commitments to safe development but do not provide data that would allow actual implementation to be verified. The analysis notes that trustworthy AI is ceasing to be only an ethical issue. It can directly affect financial performance, company value and the ability to win new customers. For publicly traded companies, investors can assess issues such as responsibility for security incidents, compliance costs, insurability of risks or dependence on computing infrastructure. Private investors are beginning to include similar questions in pre-acquisition due diligence and portfolio management. The situation is more difficult with young technology companies, which often lack a long operating history or established control procedures. The authors identify three main gaps. The first is the absence of simple evaluation procedures that translate general principles of responsible AI into concrete investment criteria. The second problem is a lack of comparable data. Investors need to know who is responsible for risk in the company, how models are tested, how suppliers are controlled and how the company handles incidents. The third gap is independent verification. Without external checks, investors have difficulty distinguishing between genuinely implemented safety practices and marketing claims. The analysis therefore highlights the importance of the OECD AI Principles, the reporting framework under the Hiroshima Process and OECD guidance on due diligence for responsible use of AI. These tools can create a common basis for assessing technology companies. The economic impact also affects enterprise customers. Companies buying AI may increasingly demand information similar to what investors seek: results of safety tests, incident records, data-access rules and independent assessments. A supplier that can document this information may gain an advantage when dealing with banks, insurers or public authorities. Inadequate documentation, by contrast, can prolong procurement approvals and increase the cost of contractual and legal reviews. The authors also stress that these are expert views published on the OECD.AI platform, not a new binding OECD position. KEY TERMS Venture capital: Investment in young or rapidly growing companies with high potential for both return and loss. Due diligence: Systematic review of business, financial, legal and technological risks before an investment or major contract. Independent verification: Review of a company’s claims and practices by an expert who is not part of its ordinary management.

OECD: 61 percent of venture capital went to AI. Investors still cannot compare companies’ safety

The full article is available to users registered at Hard Skills