Sam Altman just told Congress that superintelligence is coming, and the crypto industry is already bleeding $1.4 billion a year because AI made hacking cheap.

The Summary

  • OpenAI's CEO testified before Congress calling for urgent preparation for AGI/superintelligence risks and opportunities
  • Crypto assets worth over $1.4 billion were stolen in the past year, with AI tools dramatically lowering the barrier to exploit code vulnerabilities
  • The collision of increasingly powerful AI and immutable blockchain transactions creates a new attack surface where mistakes are permanent and theft is automated

The Signal

Altman's congressional testimony marks a rare public moment where the CEO of the leading AI lab admits we're building something we don't fully control yet. But the abstract risk of superintelligence is already materializing in concrete ways for anyone holding digital assets.

The $1.4 billion in crypto thefts isn't just another security statistic. It represents a fundamental shift in the economics of cybercrime. AI tools have commoditized exploit discovery. What used to require a team of skilled security researchers can now be automated by models that read code faster than humans and pattern-match vulnerabilities across millions of lines. The barrier to entry for sophisticated attacks just collapsed.

This matters for Web3 because smart contracts are transparent by design. Every line of code sits on-chain, readable by anyone, including AI agents trained to find the exact sequence of transactions that drains a protocol. Traditional software can patch and roll back. Blockchain transactions are final. When an AI-assisted attacker finds a flaw in a DeFi protocol managing $100 million, there's no undo button.

The irony is sharp. Web3 promised trustless, permissionless systems where code is law. AI is proving exceptionally good at finding where the law has loopholes. The same transparency that enables verification also enables automated reconnaissance at scale. We built financial infrastructure designed to be censorship-resistant and immutable, then handed increasingly capable AI the ability to read every blueprint.

The Implication

If you're building in crypto, your security model just changed. Code audits are table stakes, but they're snapshots. You need continuous monitoring systems that assume AI agents are reading your contracts 24/7 looking for edge cases. Bug bounties need to scale with the threat surface. Insurance markets for smart contract risk are about to get a lot more expensive and a lot more necessary. The race between AI-assisted defense and AI-assisted attacks is already underway. Altman's warning about superintelligence is theoretical. AI draining your protocol is not.

Source: CoinDesk