Erik Voorhees, a prominent figure in the crypto space, argued on July 28 that governments should not be allowed to decide which forms of intelligence are 'safe.' His warning comes as a growing number of tech executives push for federal oversight of advanced AI models, and as the Trump administration reportedly finalizes a voluntary testing framework.
The argument against state-defined 'safe' knowledge
Voorhees, speaking to an audience of XRP community members, said that letting governments define acceptable AI knowledge could quickly expand from banning bioweapons research to outlawing unapproved encryption and requiring AI systems to not obstruct government orders. He warned that only the United States has the cultural character to resist state control of AI knowledge, and if it fails, nowhere will it be preserved.
David Schwartz, Ripple's CTO Emeritus, expressed support for Voorhees' argument. XRP community member Bird questioned where the line on 'safe' knowledge would stop, echoing concerns about mission creep.
Industry support for federal AI testing
On the other side of the debate, several major tech leaders have called for government involvement. Google DeepMind CEO Demis Hassabis proposed a federally backed body to test and certify frontier AI models before release. OpenAI's Sam Altman called the proposal 'thoughtful,' and Microsoft's Satya Nadella welcomed it as a way to prevent serious harm from AI models.
Anthropic CEO Dario Amodei denied that his company advocated banning Chinese open-weight models, stating they support open models without dangerous capabilities. Anthropic's concerns include authoritarian governments using AI for military and surveillance purposes, as well as misuse in cyberattacks or biological threats. The company supports restrictions on advanced chip access, action against industrial-scale distillation, and safety testing for highly capable AI systems.
Coinbase CEO pushes back
Coinbase CEO Brian Armstrong rejected the idea of a federal AI testing body. He argued that existing fraud, tort, and consumer protection laws are sufficient, and questioned designing regulation around a hypothetical problem. Armstrong's stance aligns with Voorhees' skepticism of government overreach, though from a different angle.
What's at stake
The Trump administration was reportedly finalizing a framework for voluntary AI model testing by the government. That framework would not mandate compliance, but it could set a precedent for future regulation. Voorhees' warning highlights a fundamental tension: how to prevent catastrophic misuse of AI without giving the state a monopoly on acceptable knowledge.
The debate is far from settled. With the administration's voluntary framework expected soon, the question remains whether industry self-regulation and existing laws can address AI risks, or whether a new federal body is needed — and at what cost to freedom of inquiry.
