How to Avoid Debate: Scalable AI Safety via Doubly-Efficient Interactive Proofs

2026-07-03

AI Summary

This work addresses a key limitation in existing debate-based approaches to AI safety verification, which rely on the unrealistic assumption of two equally capable models with one guaranteed to be honest. To overcome this, we propose the first single-prover interactive proof framework tailored for AI alignment verification, eliminating the need for adversarial debate. Our approach is applicable in settings where an oracle—such as human judgment or an external database—is accessible. We design doubly efficient verification protocols supporting two types of oracles: computational oracles robust to a small number of query errors, and low-degree polynomial oracles. Under structured or noise-tolerant oracle access, our method enables practical and scalable verification of AI outputs, thereby transcending the constraints inherent in traditional debate-based models.

Abstract

As AI models continue to develop powerful capabilities, it becomes critical that we are able to verify that their output is aligned with our intentions. A recent line of work focuses on verification via debate, a model of interactive proofs where two competing powerful provers, or AI models, debate each other to convince a weak verifier, or a human, of the correctness of their claim. However, debate assumes that the two AI models possess equal abilities and that one of them is truthful, which may not be realistic. In this work, we show \emph{how to avoid debate}: we initiate the study of \emph{single-prover} interactive proofs for AI safety. Prior results in single-prover interactive proofs do not immediately carry over to the AI safety setting: for example, they do not work when the computation has access to an oracle, such as to human judgment or an external database such as the web. We present doubly-efficient single-prover interactive proofs and arguments for oracle-aided computations (also known as relativizing proofs), in the settings where (1) the computation is robust, in the sense that the output does not change if at most a small fraction of the answers to oracle queries are incorrect, or (2) the oracle is a low-degree polynomial. These results suggest that interactive verification is possible even without debate, under structured or noise-tolerant oracle access.

Problem

• AI safety
• interactive proofs
• single-prover
• oracle-aided computation
• verification

Innovation

• single-prover interactive proofs
• doubly-efficient
• oracle-aided computation
• AI safety
• relativizing proofs