We scan new podcasts and send you the top 5 insights daily.
Tom Milton compares model weights to chemical-industry IP. The OPCW's mass spectrometers have closed modes that return pass/fail rather than full spectra, protecting confidential formulas. He suggests AI verification needs similar systems that confirm treaty compliance without revealing weights or tokens.
Life sciences AI vendors can satisfy regulatory demands for explainability without exposing proprietary model code. By providing empirical rationale—such as matching past cohort sizes from similar indications—platforms ground recommendations while protecting their intellectual property.
As sovereign AI initiatives grow, the risk of an unchecked capabilities race increases. A shared language of evaluations could provide a "trust but verify" mechanism, akin to nuclear arms verification treaties. This allows nations to audit each other's AI for risks like runaway RSI, creating a basis for international policy.
Tom Milton argues verification is not only for outside regulators. If AI models gain nation-state-level cyber capability, labs need hardware-backed logs to confirm models have not changed their own weights or spoofed evaluation results. The same stack protects against untrusted AI as against an adversary country.
Any international or inter-company agreements to slow down AI development are impotent without the technology to verify compliance. The 'teeth' of any treaty or law will be technical verification tools, making their development a critical and urgent priority.
Emerging technologies like confidential computing offer a novel path to building trust between rivals. They could allow for mutual verification of AI systems or even sensitive government documents, with an AI providing trusted attestations (e.g., "no backdoor found") without exposing the underlying secret data.
For an AI firm, leaking source code exposes its engineering roadmap to competitors. While a major blunder, it's not a death blow because the core intellectual property—the trained model weights which represent the AI's "knowledge"—remains secure. Competitors get the blueprint, but not the trained intelligence.
Tom Milton says he is not convinced verification is strictly necessary. Nuclear deals relied partly on trust, satellites counting bombers, and spying; AI deals could also use auditors, whistleblowers, or unilateral slowdowns. But as more lab leaders and governments call for coordination, verification becomes valuable optionality.
Confidentiality concerns make AI companies hesitant to admit external human auditors into proprietary codebases and training runs. Highly capable AI auditors whose internal memory states are securely wiped immediately after compiling redaction-adjusted compliance reports can verify safety without leaking intellectual property.
Claims that AI treaties are unverifiable lack imagination. During the Cold War, the US and USSR agreed to saw bombers in half on runways, allowing spy planes to visually confirm disarmament. Similar 'outside-the-box' physical verification methods, like publicly escrowing or destroying GPUs, could work for AI.
International AI treaties are feasible. Just as nuclear arms control monitors uranium and plutonium, AI governance can monitor the choke point for advanced AI: high-end compute chips from companies like NVIDIA. Tracking the global distribution of these chips could verify compliance with development limits.