Arena AI CEO says enterprises don’t know which AI models to trust

Arena CEO Anastasios Angelopoulos has expressed concerns that companies are finding it difficult to trust AI models. He stated that enterprises have reasons to be suspicious of both frontier labs and Chinese open models. Angelopoulos noted that companies are "terrified" of working with both types of models. He believes that enterprises are struggling to determine where to build their AI future.

Arena offers a crowdsourced AI model battle royale where users pit generative AI models against each other and vote on which one performs best in a task. The company recently announced a commercial business where it sells data gathered from these battles to AI companies and businesses. However, Angelopoulos warned that frontier AI companies can change their models at any moment if they deem them unsafe.

He also highlighted that companies may be hesitant to access the full capabilities of these models due to safeguards imposed by the companies. In contrast, Chinese open model companies offer open-weight models that can be run locally, often at a lower cost. However, some may have security concerns about relying on Chinese companies.

Angelopoulos is not the only one raising these concerns, as Palantir CEO Alex Karp has also warned about the potential risks of frontier AI companies gathering valuable intellectual property and data. Nvidia CEO Jensen Huang has suggested a balanced approach, suggesting that companies use both open and proprietary models, depending on cost-effectiveness and the need to protect intellectual property.