AntSeed Builds Cryptographic Layer to Verify AI API Model Identity Claims

A research paper titled 'Real Money, Fake Models' found that some APIs claiming to serve frontier AI models were quietly delivering cheaper, lower-quality alternatives, exposing a financial incentive to deceive paying users. AntSeed, a peer-to-peer AI marketplace, is developing a verification system to address this gap, starting with cryptographic signing of responses. The signed payload records request and response hashes, buyer and seller identities, the advertised model name, and timing data, ensuring sellers cannot later deny serving specific outputs. However, the company clarifies that cryptographic attribution alone does not confirm model identity — it only proves who served which bytes. AntSeed plans to layer behavioral and statistical auditing techniques on top, including knowledge boundary probes and perturbation fingerprinting, to detect when a different or cheaper model has been substituted.