In a bold move that blends cryptocurrency with artificial intelligence, BitGo CEO Mike Belshe has publicly funded a wallet with 100 BTC and issued a direct challenge to Anthropic's AI: try to steal it. The stunt, which has captured the attention of the crypto community, is as much a security test as it is a publicity play.

The $10 Million Challenge

Belshe, who leads the digital asset custody firm BitGo, has put a significant amount of money on the line—100 Bitcoin, worth roughly $10 million at current prices. The wallet is now live on the blockchain, and the CEO has publicly dared Anthropic's most advanced AI model to find a way to access the funds.

The challenge is not just about proving a point; it's a real-world stress test for the security of self-custodied crypto assets. Belshe's move highlights the ongoing debate about whether AI-powered hacking tools could eventually pose a threat to the cryptographic keys that secure digital assets.

Why Anthropic?

Anthropic, known for its safety-first approach to AI development, has been at the forefront of creating models that are designed to be harmless and helpful. By targeting Anthropic specifically, Belshe is challenging the very principles of AI safety—asking whether even a 'safe' AI can be coerced into malicious behavior.

This isn't the first time Belshe has engaged in such public experiments. He has a history of using provocative challenges to highlight the importance of secure custody solutions. However, this particular dare has a unique twist: it pits the world's most advanced AI against the immutability of blockchain technology.

Can AI Crack Crypto Keys?

The core question at the heart of this challenge is whether AI can break the cryptographic algorithms that underpin Bitcoin. Bitcoin relies on the SHA-256 hashing algorithm and the Elliptic Curve Digital Signature Algorithm (ECDSA) for its security. These are considered computationally infeasible to crack with classical computers, even with AI assistance.

AI could potentially be used to try to guess private keys through brute force, but the keyspace is so vast (2^256 possibilities) that it would take an astronomical amount of time and energy. Even the most powerful AI supercomputers would need billions of years to exhaust all possibilities. So, on the surface, Belshe's 100 BTC seems safe.

Social Engineering and Phishing

However, AI's real threat might not come from direct cryptographic attacks, but from social engineering. AI models could be used to create highly convincing phishing scams, tricking wallet owners into revealing their private keys or sending funds to malicious addresses. Belshe's challenge could expose vulnerabilities in how humans interact with AI, rather than in the underlying technology.

Anthropic's AI, for instance, could potentially be manipulated via prompt injection attacks to reveal sensitive information or to generate malicious code. But whether it can actually steal the 100 BTC is another matter entirely.

The Community Responds

The crypto community has reacted with a mix of amusement, skepticism, and intrigue. Some see it as a clever marketing stunt that underscores the robustness of blockchain security. Others view it as a dangerous invitation, suggesting that even attempting to hack the wallet could have unintended consequences.

Security experts have pointed out that Belshe's challenge is not entirely fair. The wallet's private key is presumably in cold storage, inaccessible to any AI that doesn't have physical access. The AI would need to find a way to interact with the wallet's owner or the exchange where the funds are held, which is a different kind of security problem.

Regardless of the outcome, the challenge has sparked a valuable conversation about the intersection of AI and crypto security. It also serves as a reminder that while AI is getting smarter, the fundamental principles of cryptography remain sound.

Key Takeaways

  • Real-world security test: BitGo CEO Mike Belshe has funded a 100 BTC wallet and challenged Anthropic's AI to steal it.
  • Cryptographic resilience: Bitcoin's security is based on math that is currently infeasible to break, even with AI.
  • AI's true threat: The biggest risk may be AI-powered social engineering, not direct key cracking.
  • Community engagement: The stunt has sparked debate about AI, security, and the future of self-custody.

Whether the AI succeeds or fails, Belshe's challenge is a win for public awareness. It reminds us that the security of our digital assets depends not just on strong encryption, but also on the humans and AIs we trust with our keys.