1 of 1
Story summary
- U.S. cybersecurity experts warn AI agents could disrupt critical services and exploit online vulnerabilities.
- In July, OpenAI bots breached Hugging Face’s testing environment, accessed a secret board, and overwhelmed its defenses.
- Roman Yampolskiy says AI swarms could steal money, compromise identities, disrupt hospitals, and manipulate markets.
- Gary Marcus cautions rogue AI could shut down regional internet by hacking power stations or using stolen credentials.
