AI security startup Attestable has formally launched after raising $20 million in seed funding. The round was co-led by Altimeter Capital's Jamin Ball and TLV Partners' Yonatan Mandelbaum, with participation from Halcyon Futures, Cerca Partners and other investors. Founder Yogi said the rise of AI in critical infrastructure, national security and large enterprise systems has made trustworthy verification of how AI operates a global issue. Attestable aims to build a general verification layer for frontier AI labs, critical infrastructure and state-level use cases. The company said its system uses zero-knowledge proofs to shift trust in AI systems from data centers to mathematical verification. It said the technology can prove that a specific output was generated by an approved model, model weights, input data and execution policy without revealing model parameters or private user data, and without rerunning the model. Attestable added that it has already achieved verifiable inference for Meta Muse Glimmer 30B on a single NVIDIA H100 GPU at 85 tokens per second, producing compact proofs with quantum-resistant security that can be verified quickly. Ethereum co-founder Vitalik Buterin said the result suggests the performance overhead for zero-knowledge proofs on large language models is now close to single-digit levels, and that the next challenge is to further reduce the cost of technologies such as fully homomorphic encryption (computation on encrypted data).