Arena raises $200M and launches an AI-agent alignment index
The AI evaluation company says its new index examines unauthorized actions, false attribution and misleading completion claims in agent sessions.
Funding for an evaluation business
Arena announced a $200 million Series B at a $3.1 billion valuation on October 8. The company says Lightspeed Venture Partners and Khosla Ventures co-led the round, with several new and existing investors participating. TechCrunch noted that Arena’s January Series A valued it at $1.7 billion. Arena also says its annualized revenue has exceeded $100 million; that is the company’s run-rate claim, not a reported full-year result. The financing will support its work evaluating AI systems using interactions from its public platform and commercial evaluation products.
A narrow look at agent behavior
Alongside the round, Arena previewed an Alignment Index based on real-world agent sessions. Its initial signals flag actions beyond a user’s permission, statements falsely attributed to a user, and claims that a task was completed when the record shows otherwise. Arena says its analysis compares 27 models across 90,000 sessions and uses an AI judge with rubrics refined through human review. Those results describe Arena’s own methodology and sample. The company explicitly says the three signals cover only a small part of safety and alignment, so the index should not be read as a complete verdict on a model’s trustworthiness.