Arena AI CEO Anastasios Angelopoulos says enterprises are increasingly uncertain about which AI models to trust amid growing concerns over reliability, cost, and geopolitical risks.
Arena AI’s commercial model evaluation service has reached $100 million in annualized run-rate revenue within eight months of its launch, reflecting heightened demand for third-party assessments of AI models.
Angelopoulos, whose company operates a crowdsourced platform where users compare AI models, told Business Insider and Yahoo Finance that businesses face a “really tricky situation” when selecting AI providers. He highlighted two primary sources of concern: frontier labs such as OpenAI and Anthropic, and Chinese open-source models from companies like Moonshot and DeepSeek.
Frontier Labs Pose Reliability and Access Risks
Angelopoulos noted that while frontier AI labs offer cutting-edge models, their high costs and potential for sudden access restrictions create instability for enterprises. He pointed to recent actions by the Trump administration as an example of how model providers could be forced to shut off access without warning. Additionally, companies like Anthropic have implemented safeguards that may limit access to a model’s full capabilities, further complicating enterprise adoption.
Chinese Open Models Raise Geopolitical and Compliance Concerns
On the other hand, Angelopoulos said enterprises are also hesitant to rely on Chinese open-source models due to geopolitical tensions and regulatory uncertainty. While these models may offer cost advantages and flexibility, their association with Chinese companies introduces risks related to export controls, data sovereignty, and compliance with U.S. or international regulations.
Arena AI’s Role in the AI Evaluation Market
Arena AI’s platform allows businesses to compare AI models side-by-side through user-voted “battle royales,” providing data-driven insights into performance. The company’s commercial service, launched in September, sells aggregated evaluation data to AI developers and enterprises. This model has gained traction quickly, achieving $100 million in annualized revenue by June, underscoring the growing need for independent assessments in a crowded AI market.
Industry Response and Broader Implications
The uncertainty highlighted by Angelopoulos reflects a broader trend in the AI industry, where enterprises must balance innovation, cost, and risk management. Some companies may opt for hybrid approaches, combining models from different providers to mitigate single-point failures. Others may prioritize localized or regulated models to align with compliance requirements.
As the AI landscape evolves, the demand for transparent, reliable, and politically neutral model evaluations is expected to grow, with companies like Arena AI positioned to play a key role in guiding enterprise decisions.