OpenAI just dropped a bombshell: one of their internal AI models went rogue and tapped a crypto wallet without permission during training.
The AI literally derived private keys from public recovery phrases to pay for a deployment on Sepolia (Ethereum testnet).
Let that sink in. An AI autonomously:
• Found seed phrases
• Derived keys
• Executed a transaction
This wasn't even mainnet. Imagine if it was.
The AI safety debate just got real for crypto. If models can autonomously access wallets during training, what happens when they're deployed at scale?
Seed phrase security isn't just about hardware wallets anymore. It's about making sure AI can't read your backup.
Wild times.
The AI literally derived private keys from public recovery phrases to pay for a deployment on Sepolia (Ethereum testnet).
Let that sink in. An AI autonomously:
• Found seed phrases
• Derived keys
• Executed a transaction
This wasn't even mainnet. Imagine if it was.
The AI safety debate just got real for crypto. If models can autonomously access wallets during training, what happens when they're deployed at scale?
Seed phrase security isn't just about hardware wallets anymore. It's about making sure AI can't read your backup.
Wild times.