Anthropic says it is developing a way to verify which AI model offered a given output.

Complete the form to unlock this article and enjoy unlimited free access to all PYMNTS content — no additional logins required.

The artificial intelligence (AI) startup’s FlashProve system is designed to deliver what is known as “provable inference,” which allows users to confirm that responses came from a specific AI model, according to a Tuesday (Nov. 3) Anthropic blog post.

“When you talk to a friend in person, you have many ways to know it is really them: you can see them, hear them, shake their hand,” the post said.

“When you talk to an AI model, you have almost none. A user querying a model today cannot check that the responses came from the specific model that a developer claims is responding, as opposed to one secretly substituted in by a routing service for cost reasons, or even by an attacker who has tampered with the token stream.”

Going solely by the model’s output isn’t reliable, even if it offers up high-quality responses, Anthropic added. Cryptographic proofs of inference address this problem by allowing a model provider to commit to a model and then prove that each response came from that model.

“Such proofs could enable applications ranging from consumer products to internal security, and could contribute to a toolkit for verifying compliance with AI pacing requirements,” the post said.

Anthropic added that it believes the progress it has made in this field has important “implications for security, in pacing artificial intelligence development, in providing AI customers with integrity guarantees, in facilitating tooling for auditors and regulators, and more.”

Source link