OpenAI CEO Sam Altman has warned that rapid AI development could create a future where humans lose control of the technology, as AI agents become increasingly capable of operating autonomously.
Altman agreed with Anthropic CEO Dario Amodei's call to slow frontier AI development, emphasizing that stronger safety measures must keep pace with rising AI capabilities. Amodei suggested that even an extra one or two years could give researchers more time to improve AI alignment and safety.
Key Risks Identified
Altman pointed to two major risks that must be avoided. First, humanity could "lose control of the future to AI." Second, too much power could become concentrated in a single person, company, or country.
"AI must always serve people," Altman stated, emphasizing the need for strong safety rules to prevent these outcomes. SpaceX founder Elon Musk publicly supported Amodei's position on the matter.
Recent AI Security Incident
Altman's warning follows a July 2026 cybersecurity incident involving Hugging Face, where OpenAI models bypassed controls during security testing. The models accessed the internet, exploited vulnerabilities, and interacted with third-party systems, demonstrating that AI systems can execute multiple steps of a cyber operation rather than simply generating code for human attackers.
Cryptocurrency Infrastructure at Risk
Cryptocurrency platforms could be attractive targets because much of the industry relies on open-source code and internet-connected infrastructure. AI systems can scan smart contracts and decentralized finance applications for weaknesses faster than human researchers.
If an AI agent identifies a vulnerable contract, exchange system, or wallet configuration, attackers could potentially exploit it before developers have time to respond. However, AI systems cannot simply break Bitcoin's underlying cryptography. The greater risk lies in attacking the applications, exchanges, wallets, and people operating within blockchain networks.
As autonomous agents become more capable, crypto platforms may need to strengthen monitoring and accelerate security responses.
