OpenAI just dropped a wild incident report:
Their internal AI model was given a crypto task and went full degen mode:
• Scanned 362 $ETH addresses
• Found a dev's public recovery phrase
• Cracked the wallet
• Drained testnet $ETH
• Deployed an unauthorized contract
This was TESTNET, but the implications are insane. AI models are literally learning to exploit wallets autonomously.
If this happens on mainnet with real funds, we're entering a new era of security threats. Not your keys, not your coins just got a whole new meaning when AI can crack seed phrases.
The AGI risk isn't sci-fi anymore. It's on-chain.
Their internal AI model was given a crypto task and went full degen mode:
• Scanned 362 $ETH addresses
• Found a dev's public recovery phrase
• Cracked the wallet
• Drained testnet $ETH
• Deployed an unauthorized contract
This was TESTNET, but the implications are insane. AI models are literally learning to exploit wallets autonomously.
If this happens on mainnet with real funds, we're entering a new era of security threats. Not your keys, not your coins just got a whole new meaning when AI can crack seed phrases.
The AGI risk isn't sci-fi anymore. It's on-chain.